Docs

Strunk documentation

Last updated: August 19, 2026

Strunk is a bridge between your AI tool (Claude Code, Codex, and other MCP clients) and the doc tools your reviewers already use. You draft and revise where your repo, tools, and data context live; Strunk publishes the doc out to Google Docs for review, pulls reviewer comments back with their anchored spans, and applies approved revisions in place — replying to and resolving threads without destroying existing comments. Google Docs stays the source of truth.

Before you start

A Strunk account — sign up with your Google account, free.
A Google account, connected once so Strunk can create and update Google Docs on your behalf.
An AI tool that supports MCP. Any client that installs a remote MCP server over streamable HTTP works.

Connect Strunk

Strunk is a remote MCP server. Add it to your AI tool as a connector using the server URL below. Authentication runs over OAuth 2.0 — no API keys to paste.

https://strunk.io/mcp

Add the server URL above as a custom connector in Claude (Settings → Connectors → Add custom connector), or via your tool's MCP config.
Sign in when the OAuth window opens and approve access.
The first time you publish or open a doc, Strunk asks you to connect Google Docs. Strunk uses Google's limited drive.file permission, so it only touches docs it creates or files you explicitly open with Strunk.

The round-trip

A typical review cycle, all driven from your AI tool:

Publish. Strunk renders your markdown draft into a new Google Doc and returns its URL. Re-publishing the same doc applies a targeted diff, so existing reviewer comments and their anchors survive.
Share. Send the Google Doc to your reviewers. They comment in Google Docs — the tool they already know — with no new tool to adopt.
Pull. Strunk brings comments back into your AI tool with the exact quoted spans, resolved status, and threaded replies, so you can triage with full context.
Revise. You approve revisions in your AI tool; Strunk applies them as direct edits on the commented spans.
Reply & resolve. Strunk replies to each thread in your voice and resolves it once the comment is handled.

Tools

Strunk exposes a small set of purpose-built tools. Read-only tools run without a per-call prompt; tools that change a doc always ask first.

publish_doc — publish or re-publish a markdown draft to Google Docs.
pick_google_doc — connect an existing Google Doc so its comments can be pulled and revised.
pull_comments — read reviewer comments back with anchored spans (read-only).
inspect_doc — read the current doc structure to verify an edit landed (read-only).
apply_revisions — apply author-approved edits onto the commented spans.
reply_comment — reply to a comment thread.
resolve_comment — resolve a handled thread.
get_preferences / set_preferences — read and update reply/resolve/sharing preferences.
get_writing_rules — your writing rules, read before your AI tool drafts anything (read-only).
get_review_workflow — how your AI tool triages and answers reviewer comments (read-only).

How Strunk triages comments

Strunk never applies a substantive revision silently — your approval in your AI tool is the gate. Comments fall into three tiers: mechanical (typos, dead links, formatting — drafted and applied on your ok), judgment calls (a tradeoff you decide before anything is written in your voice), and needs context (surfaced, never fabricated). For each comment the order is reply first, then apply the edit, then resolve — so the acknowledgment stays tied to the reviewer's span even if the edit removes it.

Your data

Strunk does not store Google Doc bodies, titles, URLs, or reviewer comment text — those stay in Google Docs. It keeps only the minimal mapping needed to update the right doc. Google refresh tokens are custodied by WorkOS rather than stored in Strunk's database. See the Privacy Policy for details on what Strunk processes and how.

Support

Questions or issues? Email support@strunk.io. You can revoke Strunk's Google access at any time from your Google account settings.