Skip to content

Feature Roadmap

This doc is a high-level orientation — feature ideas, architectural direction, and historical context. Active work is tracked in GitHub Issues. When a roadmap item is scheduled, an issue is opened; the checkbox here stays as a reference.

Features are ordered by practical priority within each tier, cross-domain. Items marked with a source were identified by auditing competing projects — see decision-fork.md for full credits.

What's implemented

Sheets, Drive, Docs, and Calendar are all covered — see Tools for the full, auto-generated, per-domain list of every tool and its parameters. That list is regenerated from the source tool registry on every commit, so it's never out of sync the way a hand-maintained enumeration here would be.

Release cadence

Version Scope Signal
v0.7.0 ✅ First stable PyPI release — 63 tools across Sheets, Drive, Docs, Calendar Published 2026-06-21
v0.8.0 ✅ Tier 1 complete — all "frequently needed" items across all domains (84 tools) Published 2026-06-29
v0.8.1 Defect & documentation cleanup — no new tools, ships the QA/refactor work already on develop plus fixes for #235, #242, #213, #239, #236 Stabilizes before Tier 2 feature work begins
v0.9.0 ✅ Tier 2 complete — power-user and structured-work layer, plus defects that surfaced after v0.8.1 shipped (#248) Published 2026-09-08
v0.9.1 ✅ Security hotfix cut from v0.9.0 — download_file, download_folder, and sync_folder keep every write inside local_path. Nothing else from develop Published 2026-10-04
v0.9.2 Post-release defect fixes & infrastructure addons, plus the Gmail domain (pulled forward from v1.1+ 2026-09-23, community PR #786) Stabilizes before Tier 3 begins
v0.9.3 In flight. Feature set: Markdown → Docs fidelity (#1044 and its defects), comments as a first-class capability (breaking rename), Gmail sending safety, launch readiness. Plus a ranked fill Ships when the feature set is done and passes release QA
ankylosaurus Queued: hardening. Feature set not yet set; fill candidates moved from v0.9.3 and the decision backlog 2026-10-18 Gets a number when it goes in flight
v1.0.0 API stability declaration — Tier 3 items that make the cut + any breaking cleanups from v0.8–0.9 Backwards-compatibility commitment
v1.1.0+ Future domains — Tasks (separate minor release, needs a new API client; Gmail pulled forward to v0.9.2) Expanded scope

Tier 4 items remain backlog with no assigned version.

Queued versions are named, not numbered (#1085). A version is a named feature set plus a locked fill. Until it's in flight, its section below is headed by its name and its tickets carry a planned: <name> label, so queued versions can be reordered and a hotfix doesn't renumber them. Only the in-flight version gets a number (the next tag in sequence); at that point its planned: label becomes vX.Y.Z and its section heading becomes vX.Y.Z — <Name>. The full rules are in .claude/team-roles/kai.md §"Release scoping".

Roadmap

Tier 1 — High value, frequently needed (target: v0.8.0)

Infrastructure - [x] PyPI publish — OIDC trusted publishing (#55) — v0.7.0 stable live; uvx mcp-gee-sweet works

Sheets - [x] delete_sheet — delete a tab by name or sheetId (#115) (freema/mcp-gsheets) - [x] clear_values — clear cell content in a range without touching formatting (#116) (freema/mcp-gsheets) - [x] delete_rows / delete_columns — remove rows or columns by index (#117) (freema/mcp-gsheets) - [x] format_cells — background color, font, alignment, number format on a range (#118) (freema/mcp-gsheets) - [x] merge_cells / unmerge_cells — merge a range into one cell (#119) (freema/mcp-gsheets) - [x] freeze — freeze rows and/or columns (#120) - [x] sort_range — sort a range by one or more columns (#121)

Calendar - [x] Recurring events — RRULE support in create_event and update_event; expand_recurring in list_events; instance vs master scope documented (#155)

Docs - [x] insert_inline_image — insert an image at a document index (#145) - [x] Table structural ops — insert_table_row, delete_table_row, insert_table_column, delete_table_column (#146) - [x] create_header / create_footer — page headers and footers (#147)

Drive - [x] list_shared_with_me — files explicitly shared with the authenticated user (#135) - [x] list_recent_files — files recently accessed or modified (#136) - [x] get_storage_quota — Drive storage usage and limits (#137)

v0.8.1 — Defect & documentation cleanup (target: v0.8.1, before Tier 2 begins)

No new tools. Stabilize on what Tier 1 shipped before starting Tier 2 feature work.

Already on develop, awaiting release - [x] Playwright per-test tagging policy, conductor prompt improvements (PR #228) - [x] QA test case fixes — share_file fixtures, TC-D148 setup, Playwright tags (PR #229) - [x] SQLite cache defensive recovery from read-only/corrupted DB files at connect time (PR #230) - [x] tools/docs/__init__.py split into content.py/tables.py/style.py/layout.py (PR #232)

Remaining defects to fix before cutting the release - [x] get_sheet_data(include_grid_data=True) without a range fetches the full padded grid instead of the used range (#235) - [x] Generalize the #235 response-size safety net (post-fetch size check + configurable cap + local_path bypass) to other tools that can return large responses — export_file, get_doc_content, get_multiple_sheet_data, find_in_spreadsheet, list_file_activity. Oversized responses silently kill the MCP session today (client-side output cap, e.g. Claude Code's MAX_MCP_OUTPUT_TOKENS), requiring a server restart to recover — confirmed live, not hypothetical. Renamed MAX_GRID_DATA_RESPONSE_CHARS to MAX_TOOL_RESPONSE_CHARS (single global cap); live-verified thresholds per tool (#242) - [x] create_doc_from_file markdown $ escape renders as literal backslash+dollar in the doc — Python-Markdown's default ESCAPED_CHARS omitted $ (unlike CommonMark); fixed via a small extension adding it, respecting code-span/fenced-code protection (#213) - [x] Unknown tool parameters are silently dropped instead of raising a validation error — FastMCP's generated pydantic arg models default to ignoring extra fields, so a typo'd kwarg (e.g. parent_id instead of parent_folder_id) silently falls back to default behavior instead of erroring. Fixed centrally in the tool() decorator (applies to all ~84 tools uniformly) (#239) - [x] Auto-generate docs/tools.md from tool source as a pre-commit hook — makes the manual tools.md backfill in #236 unnecessary (PR #240) (#94) - [x] Correct docs/roadmap.md's own tool catalog counts, known-limitations.md's list_all_events entry, and rework README's/docs/client-setup.md's Configuration examples to lead with OAuth (not service account) for local/dev quick-start — currently mismatches the project's own auth priority order and PyPI's page inherits it verbatim (#236) - [x] Rewrite CONTRIBUTING.md — fix broken links to removed README anchors, fix stale "Available Tool Names" and bug-label references, add better local-dev setup examples including observability (DEBUG_LEVEL/LOG_FILE/ACCESS_LOG_FILE) (#95) - [x] Define and document community PR expectations — template, testing bar for non-tool PRs, scope/size convention, CLA/DCO/code-of-conduct decision, review turnaround (#237)

Tier 2 — Useful for structured work, plus defects surfaced since v0.8.1 (target: v0.9.0)

Defects (found after v0.8.1 shipped; too late for that release, don't warrant a standalone patch) - [x] Dev prereleases become unreachable once the base version's stable release ships — format-jinja never bumps past the last tag, so X.Y.Z.devN is a PEP 440 pre-release of the already-published X.Y.Z and always loses precedence to it; fix is bump = true in [tool.uv-dynamic-versioning] (PR #318) (#317) - [x] sync_folder doesn't recurse into subfolders — one-level-only Drive query silently reports a clean "in sync" result while ignoring every nested file (confirmed live: 225 files across 22 subfolders ignored, only the root-level file seen) (PR #328) (#315) - [x] download_folder's file loop is genuinely sequential (for f in ...: await asyncio.to_thread(...)) — the one multi-file transfer path missed when the asyncio.gather pattern was established (#183); sync_folder's own transfer step already uses it correctly (PR #351) (#316) - [x] sync_folder re-uploads every file after its first download — download never sets the local file's mtime to match Drive's modifiedTime, so any real time gap past the 5s tolerance makes the local copy look "locally newer" on the next sync; confirmed live during #315/#328 QA, reproduces on both new and long-standing files, unrelated to recursion (PR #357) (#346) - [x] Bare URLs in markdown content aren't autolinked — Python-Markdown's built-in autolink only fires on <https://...> or [text](url), not a bare URL (PR #265) (#248) - [x] zip(doc_tables, ast_tables) in emitter.py silently cross-pairs tables when one is skipped (zero-row/zero-col table), misapplying fill/merge/style requests to the wrong table — surfaced during #276's review (PR #282) (#277) - [x] Nested <li> silently deletes the parent list item's own text (HTML and Markdown paths) — handle_starttag clobbers the outer block's in-progress run buffer with no save/restore; found by Aziz auditing the markdown/HTML-to-Doc pipeline (PR #369) (#335) - [x] Nested bullet/numbered list depth is computed correctly but never emitted as Docs indentation — every list renders flat regardless of source nesting; ast_to_requests never reads BulletItem.depth; found alongside #335 — was held back until #360 merged (both rewrite the same ast_to_requests bullet-handling logic in emitter.py); #360 shipped 2026-07-24, so ready-for-development/lane-b again, queued behind #402 (#336) (PR #432) - [x] emitter.py computes Docs API insertion offsets via Python len() instead of UTF-16 code units — wrong for any astral-plane character (most emoji, some CJK/math symbols); fixed via external contributor PR #360, live-verified in PR #407 — unblocks #372's content.py split (PR #360) (#358) - [x] spreadsheet://{id}/info MCP resource throws 'FastMCP' object has no attribute 'get_lifespan_context' — discovered live during #361's QA, possibly a regression from the mcp SDK bump (#349/#350) (PR #368) (#363) - [x] Plain-text content (no wrapping tag) silently produces an empty doc body (PR #385) (#343) - [x] _get_sheet_id catches all exceptions from the sheet-lookup fetch and returns None, the same value it returns when the sheet genuinely doesn't exist — every caller (24 call sites) then misreports a transient API error as "sheet not found"; flagged during #380/#89's QA (PR #390) (#384) - [x] _AstParser.handle_endtag pops _list_ordered only on a matching close tag — a mismatched <ol>/<ul> pair leaves it permanently desynced, corrupting nesting depth for every subsequent list in the document; pre-existing, confirmed live during #369/#335's review — ready-for-development/lane-b re-added now that #401/#402 (both shipped) are no longer blocking it, queued behind #408/#409 (PR #449) (#382) - [x] When the markdown-to-Doc converter can't represent a construct (unsupported image path, thematic break ---, or anything else unhandled), it deletes that construct's entire paragraph rather than leaving an empty block — breaks fidelity with how every CommonMark viewer degrades (paragraph boundaries hold even when an inline element fails to resolve); confirmed live across 21 real policy documents converted via create_doc_from_file, every one needed manual insert_doc_text patches to restore lost paragraph breaks. General fix, not image-specific; doesn't require #333's image rendering to land first (PR #406) (#401) - [x] Paragraphs containing only whitespace/&nbsp; are silently dropped — distinct root cause from #401 (this is an ordinary, fully-understood HTML character reference, not an unsupported construct; likely an empty-after-strip check skipping emission), but same silent-collapse symptom and same real-document discovery session (PR #441) (#402) - [x] style_doc_range rejects link_url=null with "Links must include at least one type" instead of clearing an existing hyperlink — found while trying to strip a dead #slug anchor link left over from markdown conversion (see #409) (PR #446) (#408) - [x] Markdown-to-Doc conversion keeps GitHub/GitLab-style #slug heading-anchor links as dead hyperlinks instead of resolving them to Docs-native internal links (createNamedRange/bookmark targets) — links render correctly on the source Markdown host but go nowhere once converted. Depends on #408 (need working link-clearing/rewriting first). (PR #453) (#409) - [x] _get_sheet_index has the identical catch-and-swallow bug _get_sheet_id was just fixed for (#384/PR #390) — misreports transient API failures as "sheet not found" to its only caller, duplicate_sheet's insert_index default. Same fix shape as #384, same file family (sheets, not docs) as #384's own fix (PR #440) (#391) - [x] Thematic breaks (---) are silently dropped during markdown conversion with zero trace, worse than docs/design/markdown-support.md anticipated (that doc only discusses front-matter incidentally rendering as an <hr> paragraph, not vanishing entirely); confirmed resolved as a side effect of #401's general fix (PR #406) — closed as duplicate, no separate PR (#399) - [x] html_parser.py: mismatched <ol>/<ul> close can leave a stale block-interruption frame that a later, unrelated list closes by coincidence — pre-existing, confirmed live during #382's own review round, distinct from #382's own fix (PR #478) (#450) - [x] html_parser.py: whitespace-only <pre> resumed-block flush doesn't get the fresh-vs-resumed gate #402's fix gave other block tags — found during #402's own review round, pre-existing, confirmed live. Docs domain, sequence behind lane-b's current queue (PR #515) (#443) - [x] Neither the AST schema nor html_parser.py has any concept of a blockquote — <blockquote>/> quoted text silently converts to a plain, visually indistinguishable Paragraph, no indent or styling of any kind. Docs domain, same conversion-pipeline surface as #334 (lane-b's current ticket) — sequence behind it rather than run in parallel (#476) (PR #546) - [x] Inline-image pre-validation (#400) only checks Google's 25-megapixel ceiling, not its 50MB file-size ceiling — a large, low-compressibility image under the megapixel limit passes check_image_bytes silently and only fails later at the real Docs API call; auto_downscale=True also has no effect for this case, since the downscale path only triggers off a check_image_bytes error. Confirmed live during #400/PR #554's QA round 2. (#562, PR #580) - [x] list_calendar_acl doesn't handle pagination (nextPageToken) — silently returns only the first page for a calendar with enough sharing entries to paginate. Calendar domain, found during #158's review round (#460) (PR #612) - [x] add_calendar_acl silently ignores scope_value when scope_type='default' instead of rejecting it — docstring implies it's optional/ignored but the real behavior is undocumented silent acceptance of a value that has no effect; live-verified. Calendar domain, found during #158's review round (#458) (PR #616) - [x] Three convert_markdown matching/status-reporting edge cases found during PR #414's round-3 review — drive_map has no collision guard between a plain file and a converted Doc sharing the same name; _upload_local_file never stamps modifiedTime on a converted Doc, so it falls outside the sync tolerance and reports conflict instead of skipped on almost every first re-sync after upload_local_file(convert=True); a drive-only converted Doc always reports conflict even when direction='upload'. Two of the three confirmed live. ready-for-development/lane-a — scoped entirely to transfer.py, distinct from lane-b's docs-emitter defect chain, so it runs in parallel rather than queuing behind it (#422) (PR #433)

Sheets - [x] update_borders — border style, width, color on a range (PR #325) (#122) (freema/mcp-gsheets) - [x] hide_rows / hide_columns / unhide — toggle row or column visibility (#123) (freema/mcp-gsheets) (PR #311) - [x] resize_rows / resize_columns — set pixel height/width, or auto-fit to content (PR #321) (#124) (freema/mcp-gsheets) - [x] add_data_validation / get_data_validation — dropdowns, checkboxes, value constraints (#125) (freema/mcp-gsheets) (PR #361) - [x] update_sheet_properties — tab color, hide/show gridlines (PR #301) (#126) (freema/mcp-gsheets) - [x] duplicate_sheet — copy a sheet within the same spreadsheet (PR #286) (#127) - [x] Partial (rich-text) hyperlinks in update_cells (PR #380) (#89) - [x] import_csv_to_sheet — populate a spreadsheet from a local CSV file (PR #272) (#187)

Calendar - [x] create_calendar / update_calendar / delete_calendar — calendar lifecycle (PR #266) (#156) - [x] add_calendar_to_list / remove_calendar_from_list — subscribe/unsubscribe (PR #269) (#157) - [x] Calendar ACL — share a calendar with users or groups (PR #455) (#158) - [x] list_all_events — query all subscribed calendars in parallel, reusing the asyncio.gather/execute_in_thread pattern #183 established and the inline-per-item-error convention already used by get_multiple_sheet_data/share_spreadsheet/share_file (both open questions resolved; no longer decision-blocked) ready-for-development/lane-a (#194) (PR #464) - [x] Non-blocking follow-ups from #194/PR #464's review: scripts/gen_tool_docs.py's "Calendar only" SUBSETS entry omits list_all_events, shaping code duplicated with list_events, no concurrency cap on the parallel fan-out. No lane assignment yet (#466) (PR #625)

Docs - [x] insert_page_break — explicit page break at an index (#148) (PR #314) - [x] merge_table_cells — merge cells in an existing table (distinct from write-time colspan) (#150) (PR #310) - [x] Comments API — list, add, resolve doc comments (PR #324) (#151) - [x] create_named_range / create_bookmark — anchor points for internal links (PR #337) (#152) - [x] Rowspan emitter — closed as duplicate, already implemented via #100 (rowspan support) and PR #287 (nested-table colspan/rowspan) (#195) - [x] Warn/error on mixed text + nested table in same cell — shipped as the fuller Cell.children: list[Run | Table] ordered model, also fixing trailing-text-after-nested-table ordering (#275) in the same PR (PR #276) (#108) - [x] colspan/rowspan inside nested tables (PR #287) (#109) - [x] find_in_doc — search doc text and return match locations, parallel to find_in_spreadsheet; removes the manual index math currently needed to retrofit links (or any style) onto text already in a doc, raised while fixing #248 (PR #353) (#262) - [x] Markdown-to-Doc image support — ![alt](path) silently dropped today; local file path, drive:<id>, or HTTPS URL should all resolve to an inline image at the right position — deprioritized behind formatting fixes (#401/#402/#399, #403/#404) per user request; no lane assignment yet (PR #502) (#333) - [x] insert_inline_image fails with the raw Google Docs API error ("The provided image is too large") for images over ~25 megapixels — no upfront size check, no mention of the actual limit or the image's own megapixel count; diagnosing required external tooling (sips) and prior knowledge of Google's ceiling. Same image codepath as #333, which has since shipped (PR #502) — no longer blocked. (#400) (PR #554) - [x] Expose paragraph list/nesting info in get_doc_structure, plus a tool to set/change bullet membership and nesting level on an existing range (createParagraphBullets/deleteParagraphBullets equivalent) — markdown converter currently flattens indented sub-lists into the parent list with no way to detect or fix it after the fact (#334) (PR #524) - [x] Soft-break paragraph helper (single paragraph, multiple lines joined by soft breaks, explicit named style) + document delete_doc_range's paragraph-merge-inherits-neighbor-style behavior (#332) (PR #362) - [x] update_doc_from_file — update an existing Doc in place from a local .md/.html file, reading server-side like create_doc_from_file instead of round-tripping full content through the caller's context; real friction confirmed live — reporter had to fake this via delete-then-recreate. Wants #401/#402 landed first so the same paragraph-collapse defects don't just resurface via the update path (#341) (PR #564) - [x] style_doc_table_cells applies border_color/border_width/border_dash_style uniformly to all four cell edges — no per-edge control (top/bottom/left/right independently), even though the Docs API's own TableCellBorder supports it; blocks the classic "signature line" (bottom-border-only) cell pattern. Queued for lane-b behind #409/#382's remaining defect chain (defect-before-enhancement precedent) (#403) (PR #462) - [x] No way to set custom paragraph tab stops — Docs API supports tabStops on paragraphStyle directly, wrapper doesn't expose it; needed for form-style label/value column alignment without building a full table. Closed as infeasible — ParagraphStyle.tabStops is read-only in the Docs API, confirmed via live discovery-schema check (PR #471); follow-up #473 documented the borderless-table workaround (#404) - [x] Markdown export for Google Docs — get_doc_content (and/or export_file) has no markdown output, only plain-text; the write side already has a full HTML→AST→emitter markdown pipeline, but nothing mirrors it on read. Real competitive gap (taylorwilsdon's server has this). No lane assignment yet (#300) (PR #591) - [x] Extract named_ranges.py from content.py — create_named_range/create_bookmark share one private helper and have no dependency on anything else in the file; zero-risk mechanical move, do first, unblocks #339/#340 to be built directly against the new module (#371) (PR #570) - [x] Split remaining content.py responsibilities into editing.py (range-mutation tools) and images.py (image insertion), relocate the shared _collect_doc_paragraphs/_utf16_units offset helpers into ast.py, drop dead _html_to_text — sequenced after #360 merges, since that PR touches the exact helpers being relocated (#372, PR #576)

Drive - [x] restore_file / empty_trash — undelete or permanently purge trashed files (PR #386) (#138) - [x] star_file / unstar_file — mark files with a star for easy retrieval (PR #387) (#139) - [x] transfer_ownership — transfer a file to another user (PR #445) (#140) - [x] create_shortcut — create a Drive shortcut to a file (PR #405) (#141) - [x] sync_folder — option to convert markdown files to Google Docs on upload (PR #414) (#211) - [x] upload_local_file with Drive format conversion (CSV→Sheets, MD→Docs) (PR #410) (#188) - [x] Expose md5Checksum in list_files/get_file_metadata for real content-diffing, and let sync_folder diff by checksum instead of/in addition to modifiedTime — mtime is unreliable across in-place overwrites, upload_local_file's "now" stamp, and content-preserving regenerations; related to #239 (PR #472) (#274) - [x] Drive Activity API — file change history (#72) - [x] upload_local_folder doesn't route through _upload_local_file, so the convert capability (#188) is unreachable from bulk folder uploads — no way to bulk-import a folder of CSV/DOCX/PPTX into native Workspace formats. Same domain as #274, candidate for lane-a once #274 ships (PR #505) (#411) - [x] Reconcile the two independent Drive-conversion mechanisms — _upload_local_file's _CONVERT_MIME map and upload_file's convert_to_doc param do the same thing (source mimetype upload + destination-metadata mimeType override) with no cross-reference; risk of drift as more conversion paths get added. Related to #411 (#412, PR #543) - [x] sync_folder's convert_markdown upload path: two Drive files sharing the same geeSweetConvertMarkdownSource property (e.g. a Drive-UI copy of a converted Doc) collide as the same drive_map key — the later one in iteration order silently wins with no error. Narrow trigger, flagged non-blocking during PR #414's round 2 QA review (#419) - [x] sync_folder's convert_markdown upload path: create() and its metadata-only update() re-stamp share one try/except — a transient failure in the second call after a successful create() reports upload_fail even though the Doc now exists in Drive, leaving an untracked orphan with no ID recorded. Same code path as #419 — candidate for one PR fixing both (PR #645) (#420)

Infrastructure - [x] Harden concurrent-session access — fail-open on cache read/write errors, busy_timeout (PR #280) (#234) - [x] Cache reliability & configurability — runtime-configurable TTL, smarter invalidation for shared files (PR #284) (#99) - [x] Async tool execution — asyncio.gather() for parallel Google API calls; establishes the project's parallel-call pattern, which #194 should reuse rather than introducing its own ThreadPoolExecutor (PR #293) (#183) - [x] Log the running mcp-gee-sweet package version at startup — server.py::main() logs tool-filtering state but never the version, which is dynamic (uv-dynamic-versioning) so needs an importlib.metadata.version() lookup rather than a static string (PR #479) (#356) - [x] server.py::main()'s new version-logging call (#356) has no fallback for importlib.metadata.PackageNotFoundError — a broken/repackaged install or invoking main() directly without the package installed crashes the whole server before any tool-filtering log or transport startup, instead of degrading gracefully. Found during #356's own QA (PR #483) (#481) - [x] transfer_ownership missing from server://auth-status's _SA_LIMITED_TOOLS despite its own docstring documenting the OAuth-only restriction — found during #140's review. Needed a design call on generalizing _auth_status_json's failure-reason schema (storage-quota vs. no-Drive-identity are different failure classes), so routed to Joy instead of Amy. (PR #507) (#447) - [x] server://auth-status's auth_method: "adc" can't distinguish a real user credential from a service-account-backed one (GCE/Cloud Run/GKE metadata identity, or GOOGLE_APPLICATION_CREDENTIALS pointed at a key file) — both get tagged plain "adc" with can_create_in_personal_drive: true and zero limitations, which is wrong whenever ADC resolves to a service account. Found during #447's design work; not blocking it. ready-for-development (on deck, no lane yet) (#506) (PR #613) - [x] Dependabot's pip-ecosystem config doesn't always regenerate uv.lock for a version bump (confirmed on 2 of 7 PRs in the same batch) — silently defeats every --frozen CI/release install, which doesn't validate lock-vs-manifest consistency. Immediate drift fixed by hand 2026-08-06; switched to native uv ecosystem config and added a lockfile-sync CI guard (#528, PR #533) - [x] Migrate mcp.server.fastmcp → mcp.server.mcpserver (mcp SDK v2) — was held as a watch-item while v2 was alpha; mcp hit 2.0.0 stable 2026-08-20, live-confirmed fastmcp is fully removed (not deprecated) and mcpserver.MCPServer exists as expected (PR #642) (#175) - [x] Docs version switcher never renders and each release's mike deploy overwrites the previous version instead of versioning alongside it — the latest/develop split from #267 isn't actually working — config/workflow fix in PR #668; post-merge gh-pages ops (purged the stale pre-mike tree, mike set-default develop) done + live-verified: root redirects to /develop/, switcher renders, stale paths 404 (#667)

Documentation - [x] Tool-count figures have drifted out of sync across README/docs/roadmap.md/docs/tools.md (the only auto-generated one, authoritative at 122) — generate the hand-maintained ones instead of manually recounting each time this is caught. Routed to Amy (PR #550) (#467) - [x] Document transfer_ownership's server://auth-status limitation once #447's schema work lands — the tool's own docstring already promises discoverability via that resource. Delivered as part of #447's own PR (docs/auth.md's limitation bullet, decision-auth-status-limitation-categories.md, and QA's TC-I27) rather than as separate follow-up work; audited and confirmed complete, no additional doc changes needed. (PR #507) (#503) - [x] Rewrite positioning/differentiation in README and docs/index.md — official Google MCP servers are now in developer preview, which undercuts the current "no official alternative exists" framing; lead with concrete differentiators (tool breadth, stability/QA gate, design philosophy) instead, checked against what the official servers actually cover (PR #306) (#263) - [x] Markdown-support documentation gaps — README/docs/tools.md/docs/known-limitations.md/docs/design/markdown-support.md never caught up to the Docs markdown-input pipeline shipped via #102/#103/#104/#248; consolidates #297/#298/#299/#302 (PR #482) (#303) - [x] CLAUDE.md's convert_markdown note overstates its download-branch guard as reachable — round-2 review moved that case earlier in the plan-building loop, where it's now classified conflict, not failed; the guard described is dead defensive code, not live behavior. Found during PR #414's round-3 review, routed to Amy (#423) (PR #492)

v0.9.0 Full Regression QA pass (2026-09-04) — 623 PASS / 12 FAIL / 38 SKIP / 1 N/A across 674 TCs (effective 630 PASS / 5 FAIL after the two fixes below); see docs/qa/runs/v0.9.0.md for the full sign-off. Unconditional sign-off finalized in PR #703 — cleared to tag. Two real bugs found, both routed to lane B and now shipped: - [x] share_spreadsheet fails with "File not found" on any Shared Drive file — missing supportsAllDrives=True on its permissions().create() call, the only one in sharing.py without it (ready-for-development, lane-b) (#687) (PR #695) - [x] Folder-listing cache key omits max_results — silent truncation on a cache hit, and a small-limit fetch poisons the cache for later larger-limit calls (ready-for-development, lane-b) (#688) (PR #697)

v0.9.2 — Post-release defect fixes & infrastructure addons (target: v0.9.2, before Tier 3 begins)

Renumbered from v0.9.1 on 2026-10-04: v0.9.1 shipped as a security hotfix cut from v0.9.0, so this release (and the old v0.9.2, now v0.9.3) moved up one. The GitHub labels were renamed to match.

Same shape as v0.8.1 — stabilize on defects that surfaced since v0.9.0 shipped, plus a handful of infrastructure/tooling additions that don't belong to any specific tool tier. Started because v0.9.0's own scope kept growing past its original Tier 2 definition; splitting this out keeps that pattern from repeating indefinitely. The one exception to its original "no new tools" scope is the Gmail domain, pulled forward from v1.1+ on 2026-09-23 because a community contributor had already implemented it (see Gmail domain below).

Defects - [x] Isolated depth>0 bullet run renders the wrong glyph (disc instead of circle) despite correct indentation (PR #711) (#439) - [x] create_paragraph_bullets (style.py) has the same isolated-run collapse-to-nestingLevel 0 bug #439 fixed in emitter.py only — port PR #711's anchor-paragraph approach to the sibling path (PR #723) (#713) - [x] Lock in the isolated-run rendering rule — an isolated bullet run renders at its true depth (depth N → nestingLevel N) at every depth, never clamped (decided 2026-09-09: consistency over avoiding orphaned indent; the markup generator owns the depth it emits). Extend TC-DOC190 with a min_depth >= 2 case; state the rule in emitter.py's bullet-pass comment and the docs/emitter.py paragraph of CLAUDE.md (PR #731) (#714) - [x] create_doc markdown: a fenced code block whose body is raw HTML tags renders as loose plain paragraphs (fence ignored, indentation flattened) instead of a monospaced code block — python-markdown's raw-HTML-block handling appears to pre-empt fenced_code. Surfaced during #714 triage (PR #730) (#719) - [x] Bundle: Docs write side (emitter, html_parser, images, content), one PR closing #509, #567, #581, #638, #715 — lane A (PR #961) (#894) - [x] Bundle: Calendar hardening, one PR closing #459, #615, #626, #627, #726 — lane B (PR #912) (#890) - [x] delete_file(permanent=True) surfaces a misleading 404 when the caller can trash but not permanently delete (Shared Drive Content manager) — an agent can read it as "already gone" and skip cleanup; check capabilities.canDelete first — lane B (#876) (PR #888) - [x] SSE: SIGTERM with an SSE stream open logs an "Exception in ASGI application" traceback; the server still exits cleanly, pre-existing before #833 — lane A (PR #911) (#868) - [x] sync_folder's convert_markdown restamp-failure handling: quota-message gap, duplicated create()+update() try/except, test fixture isolation — companion to #420 (PR #669) (#650) - [x] sync_folder convert_markdown's modifiedTime restamp fires for all convert types (over-broad), and its create()+update() pattern is duplicated across two call sites (#435) (PR #817) - [x] _sync_level._run_one's outer catch-all leaks a raw storageQuotaExceeded message instead of the friendly _SA_QUOTA_ERROR used everywhere else — #669 follow-up, same friendly-error-consistency shape as #614/#459 (PR #710) (#670) - [x] sync_folder reports "in sync" when mtimes match but content differs (an in-place rename that preserves mtime) — use_checksum doesn't catch it either; equal-mtime + size-divergence now reports conflict for every direction (PR #712) (#659) - [x] sync_folder: an explicit use_checksum=True should verify a same-size within-tolerance pair too, overriding the abs(diff) > _SYNC_MTIME_TOLERANCE gate that #659 left in place — #712 follow-up (#716) (PR #841) - [x] list_folders with no parent hardcodes 'root' in parents, ignoring DRIVE_FOLDER_ID — wrong scope on any Shared-Drive deployment / no results for a pure service account. Found in the v0.9.0 QA pass (TC-D27), has an explicit-parent workaround (PR #717) (#689) - [x] download_file with a non-existent trailing-slash directory creates a clobbering file instead of the directory. Found in the v0.9.0 QA pass (PR #722) (#690) - [x] find_free_slots's per-calendar error shape is less detailed than list_all_events's for the same failure mode. Found in the v0.9.0 QA pass (PR #725) (#691) - [x] A bare - marker alone on its line isn't recognized as a list item by python-markdown regardless of indentation — distinct from the tracked sane_lists threshold bug. Found in the v0.9.0 QA pass; documented as a known limitation rather than fixed (PR #733) (#692) - [x] PyPI project page drops the logo — README <img src> is a relative path, which PyPI's readme_renderer strips. Found on the published v0.9.0 page (PR #707) (#706) - [x] PyPI project page 404s four README links (CONTRIBUTING.md, docs/qa/runs, docs/qa/README.md, LICENSE): repo-relative, same class as #706, which /release step 4 checks for images only. loud/narrow, in-flight by user decision since a published README can't be changed (PR #1063) (#1062) - [x] sync_folder/upload_local_file convert: Drive's modifiedTime for a Google Doc lags real edits by minutes, and the async import can overwrite the post-create restamp. Change detection for convert_markdown Docs needs a new signal. Joy scopes; decision doc in PR #819 (#814) (PR #854) - [x] sync_folder convert_markdown: a Drive-edited converted Doc stays a conflict on every run, and the only way out today is replacing the Doc (new file ID, loses sharing/comments/links). Decide: accept the Drive version as the baseline, force-upload, or both. #814 follow-up; Joy scopes (#860) (PR #1016) - [x] auth: with no token.json, a stdio server blocks on OAuth consent with no timeout, and the consent prompt goes to the protocol channel (#811) (PR #828) - [x] auth: under --transport sse, the OAuth consent wait runs synchronously on the event loop, stalling every other session and ignoring SIGTERM until consent or timeout. #811 follow-up, found in PR #828 (#833) (PR #867) - [x] Tool errors are opaque on PyPI installs: mcp 2.3 (what uvx resolves; the lock pins 2.0.0) hides any non-ToolError exception text, so #811's "run mcp-gee-sweet auth" message and uncaught Google API errors reach the client as a bare Error executing tool <name>. Observed live. Urgent: lane-a, next after #844 (#872) (PR #905) - [x] auth: offer OAuth re-authorization from the failing tool call, via URL-mode elicitation with the consent URL in a ToolError as fallback, then rebuild services in place. Covers a refresh token revoked mid-session too. No unprompted browser. Lane-a, after #872 (#873) (PR #925) - [x] sheets: an A1 range with a dangling colon (A1:, A:, 5:, :5, :B, :B5) passes _parse_a1_notation and widens through _grid_range to the whole sheet or a bigger rectangle. Hits every _grid_range_or_error tool; for sort_range/merge_cells it changes data. Found reviewing community PR #922, which rejects only a bare ":" (merged). Also covers 1:B and non-ASCII .upper() cases (see issue comment). Community PR #931 claims it (pulled from lane B 2026-10-08; back to lane B if #931 stalls) (#926) (PR 931) - [x] get_doc_as_markdown mangles inline code containing backticks. Community PR #923 (community-PR track, closes it on merge) (#936) (PR #923) - [x] get_doc_as_markdown: remaining code-span/code-block delimiter gaps (fixed `` fence, lost edge spaces, merged adjacent code runs, newlines and|inside inline code, strict-CommonMark test coverage). Same before and after #923, not a regression; found reviewing it. Same files as the #893 read-side bundle. Item 1 shipped in community PR #940; items 2–6 lane B; defect, so ahead of #893 (PR #968) ([#938](https://github.com/khuisman/mcp-gee-sweet/issues/938)) - [x] SIGTERM with an open streamable-httpGET /mcpstream truncates it ("ASGI callable returned without completing response"); sibling of #868 — lane A, after #894 (PR #975) ([#917](https://github.com/khuisman/mcp-gee-sweet/issues/917)) - [x]get_doc_as_markdown: a paragraph exported as a fenced code block drops links on its code runs and images (_render_code_blockkeeps only run text). Found reviewing #940 — lane B, after #938 (same file), ahead of #958 (PR #976) ([#951](https://github.com/khuisman/mcp-gee-sweet/issues/951)) - [x]get_doc_as_markdown: plain text ending in!right before a linked run exports as…, which renders as an image. Affects every link, not just code; fix belongs in_escape's cross-run handling. Found in PR #968 — lane B, after #951 (same file), ahead of #958 (PR #981) ([#969](https://github.com/khuisman/mcp-gee-sweet/issues/969)) - [x]get_doc_as_markdown(local_path=…)writes the JSON result envelope, not the markdown:markdown_export.pyroutes through the generic JSON-to-disk helper — lane B, after #938 (same read-side area) (PR #982) ([#958](https://github.com/khuisman/mcp-gee-sweet/issues/958)) - [x]add_chart: a cell-to-row (B2:5) or row-only (1:5) range silently charts one column —_per_column_rangesdefaultsendColumnIndexto start+1. Found reviewing #931 — lane A, after #917 (PR #983) ([#954](https://github.com/khuisman/mcp-gee-sweet/issues/954)) - [x]_parse_a1_notationhardening: one compiled grammar, index bounds, validate before the sheet-ID fetch. #931 review follow-ups, no behavior change for valid input — lane A, after #954 (same A1 area); its QA also runs #953's live checks for #931 (folded in) (PR #986) ([#955](https://github.com/khuisman/mcp-gee-sweet/issues/955)) - [x] Write side drops image links:alt(or) imports throughcreate_doc/create_doc_from_file/write_doc_contentwithout its URL, so it doesn't round-trip with #951's export. Found in PR #976 — lane A, after #955 (PR #992) ([#977](https://github.com/khuisman/mcp-gee-sweet/issues/977)) - [x] Comment quoted text comes back HTML-escaped (Café) fromadd_doc_comment,list_doc_comments, andget_doc_as_markdown(include_comments=True):comments.pypasses Drive's escapedquotedFileContent.valuethrough unchanged. Found reviewing #958 — lane B, after #959 (defect, next pick) (PR #991) ([#984](https://github.com/khuisman/mcp-gee-sweet/issues/984)) - [x]write_capped_result_to_diskwritesdir/as a file when the directory doesn't exist: a trailing separator should mean "directory" (create it, appenddefault_filename). Shared by everylocal_path` writer. Found in PR #988 — lane B, after #984 (defect) (PR #1000) (#990)

Sheets hardening (triaged out of backlog 2026-09-11 — see the "Backlog triage" note below) - [x] update_sheet_properties/dimension-tool hardening + dedup follow-ups from PR #321's review (#323) (PR #734) - [x] update_borders hardening + dedup follow-ups from PR #325's review (#330) (PR #738) - [x] No tool clears a data validation rule, and no tool exposes a range's numeric sheetId (#365) (PR #745) - [x] add_data_validation: validate the values count/shape against condition_type before the API call (#366) (PR #750) - [x] Dedupe _get_sheet_id/_get_sheet_index lookup logic; add happy-path test coverage (#442) (PR #754) - [x] _parse_a1_notation/_grid_range raise an uncaught ValueError on invalid A1 notation instead of returning {"error": ...}, across ~10 call sites in data.py/structure.py (PR #745 follow-up) (#747) (PR #756) - [x] add_data_validation's docstring and _CONDITION_VALUE_COUNTS are two independently-maintained sources of the same per-condition_type fact, no derivation link — needs a design pass (PR #750 follow-up) (#751) (PR #759) - [x] update_cells's plain-cells branch (+ get_sheet_data/get_sheet_formulas/get_multiple_sheet_data/clear_values) hand a raw range string to the Sheets values API with no local validation, leaking a raw HttpError on malformed input (PR #756 follow-up) (#757) (PR #764)

Docs hardening & dedup (triaged out of backlog 2026-09-11) - [x] html_parser.py: readability + reset-consistency nits in _emit_block_node/_block_had_unsupported_content (#417) (PR #766) - [x] emitter.py: non-blocking cleanup findings from PR #432's nested-bullet review (#434) (PR #768) - [x] html_parser: a resumed block's genuine trailing &nbsp; is still silently dropped after #402's fix (#444) (PR #772) - [x] docs/style.py/docs/content.py: no shared helper for the "omit key but keep field mask" link-clear pattern (#448) (PR #778) - [x] html_parser: merge identical ol/ul close-tag branches in handle_endtag (#451) (PR #779) - [x] Heading-anchor resolution (#409): minor duplication/efficiency cleanups in anchors.py/content.py (#454) (PR #783) - [x] insert_local_images/create_doc: image alt text is parsed but never used. Closed as infeasible — no Docs API request can write EmbeddedObject.title/description, confirmed via live discovery-schema check (#508) - [x] content.py: image-fetch-failure retry mechanism is fragile, resends the whole document per bad image (#510) (PR #787) - [x] content.py: duplicated image upload/share/revoke logic across create_doc and insert_local_images, plus an avoidable extra Drive round trip (#511) (PR #842) - [x] docs/images.py oversized-image handling: duplicated logic + unneeded full pixel decode (#400 follow-up) (#560) (PR #801) - [x] write_doc_content/update_doc_from_file leave the doc empty when the replacement fails after the clear — fixed by a pre-clear backup copy that is named in the error (backupId) and deleted on success; this replaced the 2026-10-04 revision-ID plan (docs/design/doc-replace-backup.md) — lane A, ahead of #957 (defect; same write path) (#878) (PR #1017) - [x] download_file (missing export_format) and sync_folder(dry_run=True) create local directories on calls that write nothing. Observed in Aziz's v0.9.1 hotfix regression pass; predates v0.9.1 — lane A, after #878 (defect, ahead of #957) (#1012) (PR #1023) - [x] write_doc_content/update_doc_from_file refuse the write when the pre-clear backup copy can't be created in the doc's folder (no permission to add files there, or the Shared Drive is at its item limit). Decision 2026-10-13: fall back to replacing without a backup, chosen by Drive's error reason and not the HTTP status, with a warning on success. New in v0.9.2 via #878 (from-review condition (a)) — lane A, after #1012, ahead of #957 (same write path) (#1018) (PR #1034) - [x] write_doc_content/update_doc_from_file: decide whether the cannotCopyFile (owner restricted copying for writers) and activeItemCreationLimitExceeded backup-copy refusals also replace without a backup, as #1018 does for the folder case. loud/narrow, regression since v0.9.1 (#878 introduced the copy); 2nd follow-up in #878's chain. Found in PR #1034's review — lane A, after #796. Decision 2026-10-15: keep refusing by default, with an opt-in allow_no_backup override; #1018's three reasons unchanged (#1035) (PR #1042) - [x] CLAUDE.md: fix the stale _SA_QUOTA_ERROR invariant (PR #1006 made it false) and add the GHSA-hx2f _safe_local_dest rule for Drive-name joins. CLAUDE.md-only, via Bob's gate. Kept by maintainer decision 2026-10-13 (not a review-bar ticket) — Bob, v0.9.2 (#1014) (PR #1043) - [x] export_revision: support Google Docs revisions (returned as markdown), and fix list_revisions' docstring claim that native-file revisions are pruned after 30 days. Also what #878's recovery path points callers at — lane B, after #958 (PR #988) (#959) - [x] export_revision's temporary Doc exposes an old revision's content to the default folder's audience (a whole Shared Drive) — decided 2026-10-11: under OAuth create it in My Drive root; a service account keeps the default folder, and that remaining exposure goes in docs/known-limitations.md. Found in PR #988 — lane B, after #990 (PR #1006) (#989) - [x] Quota-exceeded error tells a user identity to switch to OAuth: _quota_error_detail renders _SA_QUOTA_ERROR for every identity. Found in PR #1006's QA — v0.9.2 under condition (a): #1006 (this release) makes it reachable for an OAuth user with a full My Drive (PR #1006) (#1009) - [x] insert_local_images: temporary anyone:reader image shares aren't revoked when the call is cancelled — apply PR #877's shielded-cleanup pattern — lane B (#883) (PR #903) - [x] Bundle: Docs image-share cancellation follow-ups (from PR #903), one PR closing #904, #907, #908, #666 — lane B, moved from lane A 2026-10-13 so it runs alongside lane A's #1012 and #1018 (shares content.py with #1018, different functions) (#909) (PR #1029) - [x] _apply_doc_content: temporary anyone:reader image shares aren't revoked when the doc edit raises — revoke pass should run in a finally. Same share/revoke surface as #666 and #511 (#789) (PR #877) - [x] Extract a shared helper for the isolated-bullet-run zero-tab anchor mechanic (emitter.py ↔ style.py); harden its regression test (lane-a) (#727) (PR #735)

Drive hardening & dedup (triaged out of backlog 2026-09-11) - [x] download_folder/sync_folder progress notifications: report bytes transferred, not just file count (#352) (PR #739) - [x] download_folder's candidates tuple: fragile positional access across consumer sites, could be a dataclass (PR #739 follow-up) (#740) (PR #748) - [x] download_folder: opt-in flag to report byte-based progress/total in report_progress's structured fields (default stays file-count for consistency) (PR #739 follow-up, decision resolved) (#741) (PR #753) - [x] Unify the two progress-counter idioms in transfer.py (nonlocal vs. list-box) (#354) (PR #746) - [x] Extend notifications/progress to the remaining asyncio.gather-based multi-call tools (#355) (PR #758) - [x] sync_folder's plan/step dicts have the same untyped-access fragility #740's dataclass fix addressed for download_folder (PR #748 follow-up) (#749) (PR #755) - [x] Expose starred state via get_file_metadata/list_files — star_file/unstar_file (#139) set it but nothing reads it back (#388) (PR #760) - [x] drive: consolidate convert_markdown's duplicated matching/messaging logic in transfer.py (PR #414 follow-up) (#424) (PR #763) - [x] sync_folder: is_workspace recomputed inline in the use_checksum block instead of reusing drive_map's stashed value (#474) (PR #765) - [x] upload_local_folder: convert skip logic has a duplicate-name edge case + logic duplication vs. _upload_local_file (#514) (PR #767) - [x] sync_folder: actions list built unconditionally even when dry_run=False (#521) (PR #770) - [x] Extend the _GOOGLE_DOC_MIME sweep to remaining hardcoded-literal and comparison sites (#412 follow-up) (#544) (PR #777) - [x] refactor(drive): dedupe mime_type quote-escaping and list-files response mapping in files.py (#578) (PR #780) - [x] list_folders/list_spreadsheets no-parent + no-DRIVE_FOLDER_ID path is unbounded across all drives, no max_results (lane-a) (#718) (PR #736) - [x] download_file trailing-slash guard: fail-fast ordering, dead os.altsep clause, raw NotADirectoryError surface (PR #722 follow-up) (#724) (PR #827) - [x] upload_local_folder(convert=True): an unrelated Drive file sharing a candidate's extension-stripped name can cause a false-positive skip — decided 2026-09-23: Drive properties marker on converted uploads, legacy name-only matches still skip but are flagged as unverified (#769) (PR #800) - [x] sync_folder(convert_markdown=True): a .md name over ~95 bytes overflows Drive's 124-byte property cap — upload fails with propertyLengthLimitExceeded but Drive still creates an unmarked orphan Doc, repeated on every sync. Needs a design call on converted-Doc recognition (PR #800 follow-up) (#805) (PR #812) - [x] Deduplicate the max_results clamp expression, hand-copied 8x across drive/files.py with inconsistent caps (PR #736 follow-up) (#737) (PR #781) - [x] download_file: one pre-flight local_path check before any Drive call, for both trailing-slash and plain branches, naming the blocking path (#724/#827 follow-up; on-deck, good first issue) (PR #855) (#836) - [x] sync_folder: a failed mid-stream download leaves a truncated local file with a fresh mtime, which the next sync uploads over the intact Drive copy — download to a temp file and os.replace on success; check download_file/download_folder too. Found by code review, not yet observed live (lane-a, after #814 since both touch _run_one) (#844) (PR #884) - [x] Bundle: Drive transfer follow-ups (download_file preflight + _write_atomically), one PR closing #874, #887 — lane A (PR #1001) (#895) - [x] upload_local_file/sync_folder: compressed files upload with the inner MIME type (x.csv.gz → text/csv). Lane A, after #895 (both touch transfer.py uploads); community PR #934 closed unmerged 2026-10-10 (PR #1002) (#930) - [x] sync_folder: size_bytes never counts uploaded bytes, so an upload-only sync reports 0 (found in PR #1002's QA, observed live). Lane A, after #930 (PR #1007) (#1003) - [x] Share one HttpError reason-check helper across the storageQuotaExceeded sites. Lane A, after #1003; community PR #921 closed unmerged 2026-10-10 (PR #1011) (#916)

Testing (triaged out of backlog 2026-09-11) - [x] Bundle: test-suite tightening, one PR closing #392, #526, #551, #561, #574 — lane B (PR #924) (#891) - [x] Live verification for PR #922 (reject unbounded A1 ranges); coverage in PR #929 (PR #929) (#927) - [x] Live verification for community PR #940 (literal fences in code-block exports) (#952) (PR #1027)

Infrastructure - [x] pre-commit's ruff hook pins v0.11.9 while CI runs the project's locked 0.16.x, so a commit can pass a lint CI rejects. Seen live on PR #1000 (RUF043, three red CI runs). Moved out of the #897 bundle 2026-10-12 — lane B, after #989 (#974) (PR #1022) - [x] /release + prep-for-pr backstop checks for README render breakage on PyPI — the human-facing companion to #708's CI guard, routed to Bob (#709) (PR #967) - [x] supportsAllDrives=True is hand-inlined at ~58 Drive call sites — retrofitted twice now (#687 plus its own original miss); wants a shared wrapper or lint guard so a new call site can't omit it (#696) (PR #784) - [x] Extract a shared _BaseCache helper for the rows_fetched sufficiency check + epoch-guarded store — #688's fix copy-pasted the pattern into DriveFolderCache instead, and the epoch guard is still missing there (#698) (PR #788) - [x] Team setup: give each worktree only its own MCP server — other roles' tool names are an estimated ~15–20k of the ~80k lane bootstrap; measure first. Prioritized 2026-09-29, lane-b ahead of #789 (#850) (PR #856) - [x] Context-size backstop hook for lane sessions, plus a warning before a cold --resume — code half of #843; lane-b after #850 (#847) (PR #866) - [x] Cut CLAUDE.md and the role files down to lasting rules + pointers, moving per-PR history to docs/design//docs/decisions/ — Amy writes it (user-assigned), Bob reviews each PR under the prompt-QA gate; can land as one PR per file (PR #852, #948, #950) (#849) - [x] merge-pr.md step 6: guard the team-slot reset against in-flight work on another ticket, mirroring #837's receiving-side guard in dev.md — Bob's track (#838) (PR #853) - [x] Kai applies community at intake to outside-contributor PRs/issues, plus needs-ticket (PR with no linked issue) and verify-content (unreproduced contributor issue), and its prompt says how to resolve a community PR colliding with a lane ticket — Bob's track (PR #943) (#942) - [x] Kai intake follow-up: find a PR's issue references in comments and commit messages too, and report (don't act on) an unreferenced PR that likely duplicates a lane issue — Bob's track (PR #946) (#945) - [x] Team launch: pin each role's model and effort explicitly in the Makefile instead of inheriting the client default — Bob's track (PR #949) (#947) - [x] qa.md steps 7–8: every QA round comment starts with the fixed line ## QA round <N> (<Name>) @ <short-sha>: <sent back|approved>, so round counts and verdicts are machine-readable — Bob's track (#960) (PR #962)

Gmail domain (pulled forward from v1.1+ 2026-09-23; implementation is community PR #786, merged 2026-09-25) - [x] Gmail domain primitives (#785), via community PR #786 - [x] Gmail scopes opt-in + verify granted scopes at startup. Existing OAuth tokens weren't granted the new scopes, so refresh likely fails into a surprise browser consent or a headless hang. Must ship in the same release as #786 (#790) (PR #807) - [x] reply_to_message should honor Reply-To and handle replies to your own sent mail (#791) (PR #815) - [x] Decode message bodies using the part's charset; tolerate missing base64 padding (#792) (PR #816) - [x] get_message/get_thread return no body when Gmail delivers a large (~MB) text/plain/text/html part by attachmentId. Found in #803's live QA (#825) (PR #829) - [x] get_thread/get_message: add a body-less option — a long thread trips the response-size cap and there's no other way to get its message IDs, so it can't be read at all (#793) (PR #885) - [x] Bundle: Gmail cleanup, one PR closing #795, #802, #823 — lane B (#896) (PR #928) - [x] Local-file read policy across local-path tools — in a remote/SSE deployment these read the server's own filesystem (token, SA key), and Gmail attachments can now mail them out. Decided 2026-10-04: LOCAL_FILE_ACCESS=full|roots|off (default off under SSE/HTTP), one shared check for every local read and write; also covers _read_doc_source_file's synchronous read. Lane A after #1018, solo-only: lane B holds after #909 until it ships (#796) (PR #1039) - [x] Release-pass Gmail QA (docs/qa/tests/gmail.md, added by #786). Many cases are destructive (send/draft/trash), so they need a dedicated mailbox fixture plan — Aziz (#803) - [x] Rewrite TC-GM01–22 prescriptively, add TC-GM26–44 and the Smoke rows, per docs/qa/gmail-test-plan.md §4. Part of #803, Aziz (#822) - [x] README and docs/index.md: add Gmail coverage (recounted per-domain tool counts) and remove the comparisons with other MCP servers. Docs-only, no file overlap with PR #1039, so it runs alongside #796 despite the solo-only hold (user decision 2026-10-16) — lane B (#1033) (PR #1041)

Tool surface (from the original v1.1+ plan):

Requires gmail/v1 client and https://www.googleapis.com/auth/gmail.modify scope (or narrower gmail.readonly / gmail.send scopes where appropriate). Add gmail_service to SpreadsheetContext and wire up in auth.py lifespan.

Reading - [ ] list_messages — list messages with optional query string (same syntax as Gmail search), label filter, and pagination - [ ] get_message — fetch a single message by ID; return headers, body (plain text + HTML), and attachment metadata - [ ] list_threads — list conversation threads with optional query and label filter - [ ] get_thread — fetch all messages in a thread - [ ] list_labels — list all labels (system and user-defined)

Sending and drafts - [ ] send_message — send an email (to, cc, bcc, subject, body, optional attachments) - [ ] create_draft — create a draft without sending - [ ] send_draft — send an existing draft by ID - [ ] reply_to_message — send a reply in an existing thread

Organization - [ ] modify_labels — add or remove labels from a message or thread (covers archive, mark read/unread, star, etc.) - [ ] trash_message — move a message to trash - [ ] ~~delete_message — permanently delete a message~~ — dropped from #786 in review: needs the full https://mail.google.com/ scope

QA & docs (moved out of v0.9.0 2026-09-02 so they don't gate the release; see #629, closed) - [x] CONTRIBUTING.md: say how a ticketless PR is handled — a bug fix gets a maintainer-filed issue, a feature needs an issue explaining its reasoning first — Amy (#941) (PR #944) - [x] list_files/get_file_metadata docstrings cite a stale upload_local_file modifiedTime caveat (fixed at the root by PR #472) and carry a broken self-referential contrast — from Bob's docs/tools.md docstring sweep; Dev/QA track, batches with #682/#683 (#681) (PR 898) - [x] create_doc family docstrings name only the ~25MP inline-image limit; the ~50MB byte-size limit (#562) is equally enforced, and 3 deferring docstrings inherit the gap — from Bob's docs/tools.md docstring sweep. Routed to Amy (documentation); closed community PR #932 has a diff (#682) (PR 970) - [x] QA test-doc cross-reference cleanup after the drive.md/docs.md split (PR #582) — triaged out of backlog 2026-09-11 (#590) (PR #851) - [x] Stale "upload_local_file doesn't stamp modifiedTime" premise remains after #898; TC-D238 no longer reliably reproduces its drift. Routed to Amy (documentation); community PR #933 closed unmerged 2026-10-10, and its diff has a working TC-D238 setup. Touches transfer.py docstrings alongside lane A's #930/#916; the reworked TC-D238 needs a live run by QA (#956 closed as a duplicate) (#913) (PR 973)

Backlog triage, 2026-09-11 (Kai orchestrator session): open issues were growing ~2:1 against closures (37 opened vs. 20 closed over the prior 14 days); the 60 issues sitting in unscheduled backlog with no version were the other half of that — filed as legitimate findings, never given a real release slot. Retriaged all 60 into actual tiers rather than pruning them: everything fixing/hardening/testing already-shipped functionality went into this v0.9.1 release (now v0.9.2) (the sections above marked "triaged out of backlog"), one comments-API cleanup went to v0.9.2 (now v0.9.3) (below, since it overlaps that release's own comments-generalization work), and net-new capabilities that were never started went to Tier 3 (v1.0) or a new lighter-weight tier below Tier 3 (v1.1+) based on scope — see those sections. backlog (no version) is now empty; going forward, an issue either has a version or gets triaged promptly, it doesn't accumulate unversioned.

v0.9.3 (target: v0.9.3)

Ships ahead of the Markdown-fidelity feature set (now v0.9.4, below): what merged since v0.9.2, plus #1052 and #1124. Maintainer decision 2026-10-19.

  • [x] Blockquote indent invisible on single-line quotes (no indentFirstLine) — wrong/narrow, observed; #1044-relevant — lane B (#1056) (PR #1089)
  • [x] Stale per-thread HTTP connection fails the first call after idle (Errno 54/Errno 60), with no retry. Placed in v0.9.3 by the user 2026-10-17; re-read 2026-10-18 as severe (a call can hang for minutes), observed: urgent in-flight, exempt from the fill lock; lane A after #1057 (#1052) (PR #1126)
  • [x] get_doc_as_markdown emits a raw \u000b for a soft line break in non-code text — wrong/narrow, observed; #1044-relevant — lane A (#1057) (PR #1088)
  • [x] Publish dev and stable releases from one workflow file, so PyPI stops flagging every stable release with a "Change of provenance" warning. Added to the locked fill by the maintainer 2026-10-19; lane A after #1052; maintainer adds the new trusted publisher on PyPI before merge (#1124) (PR #1131)
  • [x] Release-scoping rules in kai.md: a version is a named feature set plus a locked 10–15-item fill, ships when the feature set is done; planned: <name> labels until in flight; feature-placement order; needs-version-call circuit breaker. Process ticket, v0.9.3 by the release-gate test: v0.9.3's fill is to be set under these rules. Kai writes it, Bob reviews (prompt-QA). The backlog re-triage is separate and maintainer-led (#1085) (PR #1106)
  • [x] Fix the stale Tier 4 "zero comment tooling on Docs today" line; close or re-scope #142 (asks for what #151 already shipped under different naming). (#662) (PR #978)
  • [x] html_parser: the </a> of an href-less anchor nested in a linked one pops the outer link, so text or an image after it is written unlinked. Found in PR #992's QA, which sent back the same bug in its new _bare_link_url stack; may close with PR #992 — lane A; moved from v0.9.2 2026-10-11 (pre-existing, invalid nested HTML only) (PR #992) (#994)
  • [x] list_spreadsheets, search_files, search_spreadsheets return trashed files — wrong/common — lane A (#1054) (PR #1083)
  • [x] Table column widths fail on any table after the first in a written doc (stale startIndex in fill_tables) — loud/narrow, observed — lane B (#1053) (PR #1082)
  • [x] style_doc_range has no font_family parameter — can't apply monospace to an existing range — community PR #1026 (#720) (PR #1026)

v0.9.4 — Markdown fidelity, comments, Gmail sending safety, launch readiness (target: v0.9.4)

Feature set and fill set by the maintainer 2026-10-18, under the release-scoping rules in #1085. The release ships when the feature set is done and passes release QA. Locked 2026-10-18, after Aziz's #1044 Phase 1 baseline: the baseline's defects joined the Markdown feature set, and #646, #957, #1078, #1100, #1103 and #1104 moved to ankylosaurus. From here only urgent fixes join (the rubric's in-flight cases and security fixes); Phase 2 findings come in only that way. The comments work follows docs/decisions/decision-comments-first-class.md (#661) and ships a breaking rename that the release notes must call out.

Renumbering, 2026-10-19/20 (maintainer): this feature set was v0.9.3. 0.9.3 now ships ahead of it (section above), so it is v0.9.4; every open ticket here moved to the v0.9.4 label on 2026-10-20, and its shipped items moved to the v0.9.3 section.

Feature set: Markdown → Docs fidelity

  • [ ] QA: Playwright-verified test plan for Markdown → Google Docs rendering (full syntax, every entry point) — absorbs #477 (full-construct fixture), closed into it 2026-10-18 (#1044)
  • [ ] Markdown sub-lists indented 2 or 3 spaces don't nest (#1071)
  • [ ] A numbered list restarts its numbering after a nested bullet list (#1072)
  • [ ] Images inside table cells are dropped silently (#1073)
  • [ ] HTML input: whitespace between list tags becomes stray bullet paragraphs (#1074)
  • [ ] A bare '-' (empty list item) isn't recognized in Markdown (#1075)
  • [ ] Markdown strikethrough (~~text~~) and HTML aren't converted (#1077)
  • [ ] get_doc_as_markdown indents a numbered sub-list under a numbered item by only 2 spaces, so it re-imports flat. Found in the 0.9.3 release-QA pass (#1128)
  • [ ] search_files/search_spreadsheets/list_spreadsheets miss Shared Drive files under a service account when no folder is given. wrong/narrow, observed in the 0.9.3 release-QA pass; v0.9.3's fill is locked (#1127)
  • [ ] Loose list items (

  • …

    • ) lose their bullet — lane A, html_parser cluster (#1111)
  • [ ] Markdown soft line breaks become paragraph breaks — html_parser cluster (#1093)
  • [ ] Write side:
    in a prose paragraph becomes a paragraph break, not a soft break — html_parser cluster (#1091)
  • [ ] A fenced code block or second paragraph inside a list item breaks the list — html_parser cluster; fence half may fold into #1112 (#1094)
  • [ ] Adjacent blocks run together: separate quotes merge, consecutive code blocks merge, quotes have no spacing — lane B, after #1056 (#1098)
  • [ ] A thematic break (---) renders as an empty paragraph (#1096)
  • [ ] GFM table column alignment is ignored (#1095)
  • [ ] A table takes on the style of the paragraph that follows it (#1097)
  • [ ] An image the Docs API can't fetch (e.g. an SVG badge) disappears without a trace (#1099)
  • [ ] Drive's native Markdown import (EP-5 to EP-7) renders differently from our converter — v0.9.3 scope: document EP-5 to EP-7 as Drive's engine; routing is #1113 (#1101)
  • [ ] upload_file(source_format="markdown", convert_to_doc=True) (EP-8) loses list nesting, code styling and task glyphs, and is nondeterministic — route EP-8 through the create_doc Markdown path, plus a CHANGELOG line (#1102)
  • [ ] Spike: replace Python-Markdown with markdown-it-py for Markdown → Docs — Joy spike; if yes, one swap ticket replaces #1071, #1075, #1077 and #1094's fence half (#1112)

Feature set: comments

  • [ ] Generalize the three comment tools to tools/drive/comments.py and rename list_doc_comments/add_doc_comment/resolve_doc_comment → list_file_comments/add_file_comment/resolve_comment (param doc_id → file_id); retire tools/docs/comments.py. Breaking change. (#663)
  • [ ] Full comments/replies CRUD parity — get/update/delete for comments; plain reply/reopen/update/delete/list for replies. Depends on #663. (#664)
  • [ ] Follow-up cleanup for the docs comments API (add_doc_comment/resolve_doc_comment/list_doc_comments) — same surface #663/#664 generalize, do it there rather than twice; triaged out of backlog 2026-09-11 (#329)

Feature set: Gmail sending safety

  • [ ] Safe-by-default sending: send_message/reply_to_message/send_draft don't deliver without explicit confirmation (drafts-first or elicitation), flag external-domain and over-threshold recipients, documented opt-out. Behavior change, needs a CHANGELOG.md note. Held from lanes until a review of existing practice settles the mechanism, thresholds and opt-out (#1081)

Feature set: launch readiness

  • [ ] Ship an MCP Bundle (.mcpb) for one-click install and Smithery listing (#1036)
  • [ ] Demo screenshot or GIF for the README — routed to Amy; how the asset gets produced is still open (#589)
  • [ ] Fact-check and word competitive/benchmark claims once real benchmark numbers exist — routed to Amy, triaged out of backlog 2026-09-11 — only if the competitor comparisons come back; otherwise it moves out (#585)

Fill (rank order, locked 2026-10-18)

  • [ ] transfer.py interpolates folder IDs unescaped into Drive queries (#1084)
  • [ ] get_doc_structure hides inline images, so an image's index can't be found (#1070)
  • [ ] create_paragraph_bullets adds the requested nesting level to an item's current level instead of setting it (#1069)
  • [ ] transfer_ownership(send_notification=False) can never succeed — loud/narrow (#1055)
  • [ ] create_bookmark doesn't create a bookmark Docs lists — docstring fix (option 2) now (#1076)
  • [ ] Docstrings, one item: Tool docstrings: strip internal issue refs again, and add a guard test (#1045), Docstrings: 'Requires OAuth or ADC auth' is wrong for a service account on a Shared Drive (#1046), list_spreadsheets docstring sends callers to list_files for a result cap it now has (#1047), Docstring fixes: named-style resets, and set_cache_ttl's transport claim (#1058), Docstrings: create_paragraph_bullets carries internal API-quirk mechanics plus residual bare issue refs in style.py tool docstrings (#683)
  • [ ] font_family, one item: Document font_family on insert_softbreak_paragraph and widen the style_doc_range font test (#1037), Harden font_family in Docs text-style requests: preserve weight, reject empty values, share one builder (#1038)
  • [ ] Refresh stale TC expectations found in the v0.9.2 pass (#1060)
  • [ ] Bundle: CI tidy (on deck, no lane), one PR closing #536, #619, #308, #708 (#897)

Moved out 2026-10-18: #379 (unversioned, Joy decides whether to do it); #378 closed (ASYNC240 stays off). Everything else formerly here is under ankylosaurus below.

ankylosaurus — hardening (queued, label planned: ankylosaurus; gets a number when it goes in flight)

Fill candidates moved from v0.9.3 and the decision-needed backlog on 2026-10-18. The maintainer sets this version's feature set and locks its fill when it opens.

Decided behavior changes

  • [ ] batch_update_cells empty input — validation error vs silent no-op — decided 2026-10-18: return {"error": ...} (#32)
  • [ ] search_spreadsheets empty query — return all or require non-empty input — decided 2026-10-18: return {"error": ...} pointing to list_spreadsheets (#39)
  • [ ] Deprecate share_spreadsheet in favor of share_file, and remove it before 1.0 (decided 2026-10-18) (#665)

Candidates

  • [ ] Optional expiration on sharing tools (#1059)
  • [ ] server://auth-status doesn't list export_revision under a service account's no_drive_storage_quota limits. Found in PR #1006's QA; predates v0.9.2 and not observed failing, so v0.9.3 per the review-ticket rule (#1010)
  • [ ] _parse_a1_notation rejects A:B5, which the Sheets values API accepts as A5:B — decided 2026-10-11: keep rejecting (a destructive GridRange tool acting on "row 5 onward" is worse than an error); document in the docstring that this is deliberately stricter than Sheets. Found in PR #986 — lane A; moved from v0.9.2 2026-10-11 (docstring only, current behavior is the safe one) (#987)
  • [ ] Bundle: Calendar follow-ups from PR #912, one PR closing #915, #918 — lane B, after #892 (#985)
  • [ ] Bundle: Docs read side (doc_to_ast/ast_to_markdown), one PR closing #596, #600, #732 — lane B (#893)
  • [ ] Bundle: server/infra hygiene (start after #872), one PR closing #517, #607, #614, #579, #637 — lane B (#892)
  • [ ] Dedicated QA calendar, teardown for creation-only tests, and a cleanup_stray_artifacts sweep — split from #304, no second account needed — Aziz (#886)
  • [ ] sync_folder(export_format=...): converted Sheets/Slides probably have #814's modifiedTime lag and restamp race too. Suspected, not yet observed (#818)
  • [ ] Add a tool to set (and read back) Google Docs table column widths — tableColumnProperties isn't reachable through any tool today (#809)
  • [ ] create_spreadsheet/import_csv_to_sheet response has no web link, unlike sibling create/copy tools — enhancement or fix TC-D01's expectation instead, still open. Found in the v0.9.0 QA pass (#693)
  • [ ] Several drive_files.md cases assume My-Drive semantics now that fixtures live in a Shared Drive (TC-D58/D61/D03/D59/D27/D04/D175); TC-D27 overlaps #689's list_folders gap — Aziz reviews during Full Regression prep (#680)
  • [ ] QA .env has no single source of truth — setup_team.sh only copies .env into a worktree when absent, so fixture-ID changes never propagate (had to hand-sync 9 copies for #305); also run.md vs setup.md disagree on docs/qa/.env vs repo-root .env. Surfaced by PR #677, relates to #648 (#679)
  • [ ] TC-DOC80 can no longer trip the response-size cap post-#519 — large-doc fixture is under the raised default; needs a per-case low-cap override — Aziz (#678)
  • [ ] Automate the mechanical release.md steps as GitHub Actions (#660)
  • [ ] Harden tests/integration's local-fs live QA harness — error surfacing on API failures, subprocess/handshake timeouts, cleanup-masking a real test failure, per-test subprocess overhead (#648)
  • [ ] Define the correctness rubric and own execution for the composite-vs-primitive efficiency benchmark — Aziz (#584)
  • [ ] qa: search_files has no unit test coverage for the error-dict path (#573)
  • [ ] Investigate other MCP clients' response-overflow handling for platform-specific MAX_TOOL_RESPONSE_CHARS guidance (#555)
  • [ ] Automated tooling for dependency-bump security review — today's manual skim can miss a well-disguised supply-chain compromise (#520)
  • [ ] Split tools/sheets/structure.py by domain (1865 lines, 25 tools) (#376)
  • [ ] Domain and public sharing tests (TC-D135–D139) — decision landed 2026-09-02: provision a real non-Google test email for external-share coverage; the type=anyone public-link tests stay unresolved (exposure risk not yet accepted) (#49)

Moved from v0.9.3 at the lock, 2026-10-18

  • [ ] update_doc_from_file: opt-in incremental (mode="patch") update that diffs block by block and rewrites only changed blocks, so comment anchors and revision history survive — lane A, after #878 (same write-path files) (#957)
  • [ ] Interaction-log middleware for tool calls — structured, append-only JSONL log per call (inputs, duration, cache hit, error), opt-in ID redaction, swappable backend (#646)
  • [ ] get_doc_as_markdown: nested-table placeholder runs into the cell text (#1078)
  • [ ] A document ending in a table can get a blank last page (#1100)
  • [ ] get_doc_as_markdown exports the collapsed pre-table paragraph as an empty heading (#1103)
  • [ ] get_doc_as_markdown exports Drive-imported lists as blockquoted numbered items (#1104)
  • [ ] Design: route Drive's Markdown import paths (EP-5 to EP-7) through our converter — Joy design; EP-5 to EP-7 stay on Drive's engine in v0.9.3 (#1113)

Triaged in 2026-10-18

  • [ ] create_paragraph_bullets drops a blockquoted paragraph's quote indent — wrong/narrow, observed; from PR #1089's review, after v0.9.3's lock (#1090)
  • [ ] DocContentCache.store has no epoch guard; a pre-edit fetch can overwrite an edit's mark_dirty — wrong/rare (#1032)
  • [ ] Doc-edit tools outside content/images leave the doc cache stale when cancelled mid-edit — wrong/rare (#1030)
  • [ ] pre-commit's ruff-format hook skips Markdown, but CI's ruff format --check . formats it (#1087)
  • [ ] Make scripts/qa_compile_results.py the standard compile step of a release QA pass (#1067)
  • [ ] Track release-QA spend across releases so passes can be sized ahead of time (#1068)

Tier 3 — Advanced / occasionally needed (target: v1.0.0)

Sheets - [ ] add_conditional_formatting / delete_conditional_formatting (#129) (freema/mcp-gsheets) - [ ] add_named_range / delete_named_range (#130) (piotr-agier/google-drive-mcp) - [ ] protect_sheet / protect_range — lock a sheet or range against edits (#128) (piotr-agier/google-drive-mcp) - [ ] set_basic_filter / clear_basic_filter — toggle autofilter on a range (#196) (freema/mcp-gsheets) - [ ] get_sheet_dimensions — read column widths, row heights, frozen counts; triaged out of backlog 2026-09-11 (#132) (freema/mcp-gsheets) - [ ] add_note / clear_note — cell notes (distinct from comments); triaged out of backlog 2026-09-11 (#131) (freema/mcp-gsheets) - [ ] Developer metadata — key-value metadata attached to rows, columns, or ranges; triaged out of backlog 2026-09-11 (#134)

Calendar - [ ] quick_add — create an event from a natural language string (#159) - [ ] Event color coding — colorId on create/update (#160) - [ ] Conference data — auto-create Google Meet link (#161) - [ ] Event attachments — link Drive files to events (#162) - [ ] Reminders — per-event override reminders (email, popup, minutes before) (#163) - [ ] Working location / OOO / Focus time events; triaged out of backlog 2026-09-11 (#164)

Docs - [ ] create_footnote — footnote at a run index (#149) - [ ] create_table_of_contents (#153) - [ ] insert_section_break — section breaks for per-section column/margin layout (#154) - [ ] Raw batchUpdate passthrough for Docs (escape hatch, mirroring Sheets' batch_update); triaged out of backlog 2026-09-11 (#339) - [ ] Read/list/delete tools for named ranges (asymmetric with existing create_named_range); triaged out of backlog 2026-09-11 (#340)

Drive - [x] File comments — list and add comments on Drive files — closed as superseded: list_doc_comments/add_doc_comment (#151, PR #324) already call the generic Drive comments resource and work on any file type; the rename to file-generic names is #663 and the rest of CRUD is #664, both v0.9.3 (#142) - [ ] Watch notifications — webhook push on file changes (#143) - [ ] Labels API — custom metadata labels on Drive files (#144) - [ ] empty_trash — optional fan-out across all accessible Shared Drives in one call, instead of one call per drive; considered and deliberately deferred during #386's review (#389)

Gmail - [ ] Service-account Gmail via domain-wide delegation (with_subject() in _service_account_creds) — Gmail is OAuth-only until then (#794)

Tier 3.5 — Planned but bigger-lift (target: v1.1+)

Net-new capabilities substantial enough (new subsystem, external infra like webhooks) to warrant their own design pass rather than a single PR — later than Tier 3, still a real target rather than an indefinitely-parked wishlist item.

  • [ ] Pivot tables — create and update pivot table specs via batchUpdate; triaged out of backlog 2026-09-11 (#133)
  • [ ] Calendar watch notifications — webhook push on calendar or event changes; triaged out of backlog 2026-09-11 (#165)
  • [ ] get_doc_as_markdown: parallelize doc fetch and comments fetch (#595)

Tier 4 — Nice to have / niche (no assigned version)

Sheets - [ ] Conditional formatting rules — add/get/delete (a-bonus/google-docs-mcp) - [ ] Protected ranges — protect/unprotect a range (a-bonus/google-docs-mcp) - [ ] Dropdown / data validation on a range (a-bonus/google-docs-mcp) - [ ] Row grouping (outline/collapse) — group/ungroup rows (a-bonus/google-docs-mcp) - [ ] Auto-resize columns/rows to fit content (a-bonus/google-docs-mcp) - [ ] Write-side column widths / row heights — companion to the read-only get_sheet_dimensions above (#132) (a-bonus/google-docs-mcp) - [ ] Native Sheets "Tables" object support (create/list/get/delete/update a structured Table, distinct from a plain range) — a newer Sheets API feature we don't touch at all (a-bonus/google-docs-mcp) - [ ] Copy formatting only, without values (paste-special-style) (a-bonus/google-docs-mcp)

Docs - [ ] Multi-tab document support (list/add/rename tabs) — Google Docs' newer per-document tabs feature (a-bonus/google-docs-mcp) - [ ] Smart chips — date chips, person chips, rich links; list_smart_chips to enumerate existing ones (a-bonus/google-docs-mcp) - [ ] Section breaks and per-section page styling (a-bonus/google-docs-mcp) - [ ] Page breaks (a-bonus/google-docs-mcp) - [ ] Table structure discovery and cloning (read a table's row/column structure; clone an existing table) (a-bonus/google-docs-mcp) - [ ] Finer-grained table styling — per-border, per-column-width, per-row-height table styling beyond what style_doc_table_cells currently exposes (a-bonus/google-docs-mcp) - [ ] Footnote management (piotr-agier/google-drive-mcp)

Infrastructure - [x] notifications/progress feedback for long-running transfer calls (download_folder, sync_folder) — no tool uses MCP progress reporting today, would establish a new pattern; split from #316 (PR #351) (#319)

Decisions Needed — blocking backlog (no assigned version)

All eleven Phase 1 QA decisions were settled 2026-10-18: #31, #37 and #40 were already done in code; #33, #34, #35, #36, #38 and #78 keep current behavior; #32 and #39 moved to ankylosaurus as small fixes. Nothing is open here.

Bugs found in Phase 1 QA run (2026-06-02)

add_chart — multi-column range fails (BUG-1)

Severity: High — affects all practical chart use cases. The tool passes the full data range (e.g. A1:D5) as a single ChartSourceRange. The Sheets API requires separate source range objects per column — one for the domain (X axis / categories) and one per series. Fix: parse the A1 range, split into domain (first column) and series (remaining columns), pass as separate entries in sources. ✅ Fixed.

add_chart HISTOGRAM — wrong API spec (BUG-2)

Severity: Medium — HISTOGRAM specifically broken even after BUG-1 fixed. HISTOGRAM is not a valid BasicChartType enum value. The Sheets API uses a histogramChart spec field rather than basicChart.chartType. Fix: detect chart_type == "HISTOGRAM" and build a histogramChart spec instead. ✅ Fixed.

add_chart BAR — wrong series target axis (BUG-3)

Severity: Medium — BAR charts fail with "series may only target the BOTTOM_AXIS". BAR charts (horizontal) have a horizontal value axis. Series must target BOTTOM_AXIS, not LEFT_AXIS. Fix: detect chart_type == "BAR" and set targetAxis accordingly. ✅ Fixed.

add_chart COMBO — no per-series type specified (BUG-4)

Severity: Medium — COMBO charts fail with "No basic chart type specified". The Sheets API requires each series in a COMBO chart to declare its own type. Fix: when chart_type == "COMBO", add type: COLUMN to all series except the last, which gets type: LINE. ✅ Fixed.

TC-W03 — test case assumption wrong

The test expected the API to silently truncate a 2D array that's wider than the target range. The API actually rejects it with a 400 error. The test case needs to be updated to reflect correct API behaviour. ✅ Fixed.

Testing

Unit tests (456 passing as of 2026-06-24)

  • [x] Add pytest and pytest-cov as dev dependencies
  • [x] Cache logic — TTL expiry, dirty flag, partial invalidation for all five cache classes; in-memory SQLite
  • [x] A1 notation helpers — _parse_a1_notation, _column_index_to_letter, _letter_to_column_index
  • [x] Tool filtering — tools excluded when not in ENABLED_TOOLS
  • [x] Service account quota error handling — structured error on 403 quota exceeded; non-quota 403s raise
  • [x] server://auth-status resource — correct capabilities for service_account, oauth, and adc
  • [x] Docs AST pipeline — html_to_ast, ast_to_requests, fill_tables, nested table emitter, cell style emitter, run style Phase 3 fields
  • [x] Docs tools — write_doc_content, style_doc_range, style_doc_table_cells, get_doc_theme, get_doc_named_styles, apply_theme
  • [x] Fix BUG-1–4: add_chart multi-column ranges, HISTOGRAM spec, BAR axis, COMBO per-series types
  • [x] Update TC-W03 test case — API rejects oversized 2D arrays, does not silently truncate

Live QA (AI-driven test cases in docs/qa/tests/)

  • [x] Sheets read, write, management, charts — sheets_read.md, sheets_write.md, sheets_mgmt.md, sheets_charts.md
  • [x] Drive files and sharing — drive.md
  • [x] Google Docs tools (all 13) — docs.md
  • [x] Calendar tools — calendar.md
  • [x] Infrastructure — infra.md
  • [ ] QA gaps targeted for v0.9 — disposition of v0.8.0 SKIP entries, done, triaged the rest below ([x] #227, PR #493); dedicated QA Google account + calendar + pollution root-cause fixes, consolidates #225/#226/#249 — deferred out of v0.9.0 (2026-09-02, paid-account cost not justified for the volume); revisit via alternate OAuth client configs; split 2026-10-04: multi-user testing stays here at v1.0, pollution cleanup moved to #886 (v0.9.3) (#304); second Drive folder + shared drive fixtures, consolidates #44/#48 — done as a full fixture-set re-provision into the mcp-gee-sweet-shared Shared Drive (Workspace-org migration overtook the original scope); TEST_FOLDER_2_ID half retired, TC-D123 permanently skipped ([x] #305, PR #677); two Aziz-routed test-doc fixups surfaced by PR #677's fixture re-provisioning, both to be handled during Full Regression pass prep — TC-DOC80 can no longer trip the response cap post-#519, needs a per-case low-cap override (#678); several drive_files.md cases assume My-Drive semantics now that fixtures live in a Shared Drive (TC-D58/D61/D03/D59/D27/D04/D175), and TC-D27 may expose a real list_folders Shared-Drive tool gap (#680); domain/public sharing coverage — moved to v0.9.1 (2026-09-02) (#49); local-filesystem coverage ([x] #50, PR #644); environment-constraint coverage — closed 2026-09-02 as duplicate, untangled into #305 (shared drives) / #49 (external + public sharing) / #304 (second account, deferred) / #673 (concurrent-session procedure) ([x] #53); concurrent-session test procedure — 2 subagents + mkdir barrier against a main-checkout server, split from #53 — procedure + pre-approved-SKIP retirement via PR #675, aziz.md release-pass wiring via PR #674; live TC-I24/TC-I02/TC-R36 runs happen in the v0.9.0 Full Regression pass (tracked in docs/qa/runs/v0.9.0.md), not a standalone issue ([x] #673, PRs #674/#675); split drive.md/docs.md test files by submodule, do before #264/#224 to avoid a file-rename conflict ([x] #233, PR #582); Playwright tag consistency audit + formalize the required/spot-check/skip tiers, raised while fixing #248, sequenced after #233 ([x] #264, PR #592, #598); image fixtures for insert_inline_image, sequenced after #233 ([x] #224, PR #639); dedicated screenshots folder, independent of the rest ([x] #231, PR #583); add_chart test cases leave charts behind with no teardown, polluting the shared Sales fixture across every QA run — same pollution-cluster shape as #304/#305 ([x] #322, PR #633); plus 7 small unit-test-coverage gaps filed 2026-08-02/08, each self-contained/ad hoc pickup — [x] #487 (PR #559), [x] #488 (PR #563), [x] #489 (PR #569), [x] #490 (PR #572), [x] #491 (PR #575), [x] #494 (PR #577), [x] #495 (PR #634)
  • See GitHub Issues (label: qa) for open QA gaps and fixture issues

Infrastructure / internal

  • [x] Migrate cache persistence — replaced four /tmp/*.json files with a single SQLite DB (/tmp/mcp_gee_sweet.db, configurable via CACHE_DB_PATH); one table, four namespaces; WAL mode
  • [x] Open PR to xing5 from upstream-observability branch (structured logging, per-tool timing, cache_discovery=False) — PR #79
  • [x] Fork repo and rename to mcp-gee-sweet; README credits xing5, freema, and piotr-agier
  • [x] PyPI publish — v0.7.0 stable on PyPI; dev track publishes 0.7.0.devN on every push to develop
  • [ ] Live QA system — fixture setup, per-release run files (docs/qa/runs/vX.Y.Z.md), Playwright OAuth automation (#173) (gate for v0.8.0)
  • See GitHub Issues (label: infrastructure) for open infrastructure work

Future domains

Tasks

Requires tasks/v1 client and https://www.googleapis.com/auth/tasks scope. Add tasks_service to SpreadsheetContext and wire up in auth.py lifespan alongside the existing clients.

Task lists - [ ] list_task_lists — list all task lists for the authenticated user - [ ] get_task_list — fetch metadata for a single task list - [ ] create_task_list — create a new task list - [ ] delete_task_list — delete a task list and all its tasks

Tasks - [ ] list_tasks — list tasks in a task list with optional due date filter and completed/hidden flags - [ ] get_task — fetch a single task by task list ID + task ID - [ ] create_task — create a task (title, notes, due date, parent for subtasks) - [ ] update_task — update fields on an existing task (tasks().patch()) - [ ] delete_task — delete a task - [ ] complete_task — mark a task as completed (shortcut for update_task with status='completed') - [ ] clear_completed — delete all completed tasks from a list (tasks().clear())

Potential / under consideration

  • Google Keep — philosophically in scope (Workspace productivity tool) but the Keep API v1 is read-only for most operations and was historically restricted to Workspace Business/Enterprise accounts. Creating and editing notes via an officially supported third-party API is not currently possible. Revisit if Google opens the API further.

  • SQLite cache encryption at rest — the cache DB (/tmp/mcp_gee_sweet.db) stores Google Sheets data in plaintext. For deployments that handle sensitive data, consider SQLCipher (open-source, AES-256, mostly API-compatible with standard SQLite) or rely on filesystem-level encryption (FileVault, LUKS, BitLocker). The official SQLite Encryption Extension (SEE) is an alternative but is commercial/proprietary. Not needed for typical local-dev use.

Inspiration and credits

  • xing5/mcp-google-sheets — original upstream this project was forked from
  • freema/mcp-gsheets — most comprehensive Sheets-specific MCP server; primary source for formatting, validation, and sheet property roadmap items
  • piotr-agier/google-drive-mcp — full Workspace suite; primary source for Drive file operations, permissions, and named/protected range roadmap items
  • a-bonus/google-docs-mcp — source-verified ~44 Sheets tools / ~39 Docs tools (one-file-per-operation style, exceeds its own README's claimed counts); primary source for the Sheets comments/conditional-formatting/protected-ranges/validation/grouping cluster and the Docs comments/tabs/smart-chips/section-breaks cluster added 2026-07-14
  • taylorwilsdon/google_workspace_mcp — 12-service, source-verified 115-tool server; only project reviewed with a confirmed Markdown export tool (get_doc_as_markdown, see #300) — our Markdown support is currently input-only