Independent Editorial Reviewer
Reads approved input files by ID from approved Google Drive folders and saves review reports to a separate Drive folder; report folders are refused as inputs.
Provides independent editorial review by sending drafts, documents, illustrations, scripts, or comics to OpenAI and returning structured findings for critique, visual QA, comic lettering checks, and whole-work reviews.
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
Independent Editorial Reviewer
An MCP server on Cloudflare Workers that gives Claude a second opinion from a different model. Claude sends text, an illustration or a whole script or comic; OpenAI reviews it independently; Claude gets back structured findings it didn't write itself.
It's a personal tool, built and run for one owner's creative projects, and published as an example of a small, production-run AI service: paid calls behind budgets, sign-in behind OAuth, and deploys only from a reviewed commit.
Production: 5.0.0-clean-workspace.1 (the earlier review engine, v4.1.0, is kept unchanged inside it).
Tools
MCP tool | What it does | Cost |
| Critique of a document, draft or plan | Paid |
| Visual QA of a page or illustration against a script and a checklist | Paid |
| Six checks on speech balloons: dialogue, speaker, tails, reading order, placement, legibility | Paid |
| Starts a whole-work review of a script, an illustrated PDF, or two revisions, as a background job | Paid, within a per-job and daily cap |
| Reads a job's status and findings | Free |
The System Guide has the inputs, outputs, the 13 editorial criteria, cost controls and source rules.
Related MCP server: OpenClaw MCP Server
How it fits together
Claude ──OAuth 2.1──▶ Worker (/mcp) ──▶ OpenAI
▲ │
│ ├─ Durable Objects: job state, daily budget reservations
sign-in behind └─ Google Drive: approved input folders, saved reports
Cloudflare AccessSign-in: claude.ai connects with OAuth 2.1 (dynamic client registration, PKCE) through
@cloudflare/workers-oauth-provider. The consent page sits behind Cloudflare Access, and the Worker checks the Access token itself (signature, audience, issuer, expiry and the one allowed email).One door: after the OAuth check, the request goes to the reviewer with its inner key set server-side, so callers never hold it. Only
/mcp,/healthand the OAuth endpoints exist; there's noworkers.devaddress.Money: whole-work reviews reserve an estimated cost against a per-job and daily cap before they start. The three older tools are paid per call and sit outside that limiter (a known gap).
Sources: the Worker reads only files explicitly passed by ID, and only from approved Drive folders, checked by parent before download. Reports are written to their own folder and refused as inputs.
Built and shipped
Tests run on every push in GitHub Actions: 63 unit tests, two runtime tests on the real bundle (the reviewer, and the sign-in end to end), a real-browser test of the consent page, and a check that the older engine is byte-for-byte as reviewed. See VALIDATION.md.
Deploys only from a commit the owner names, staging first, through
scripts/deploy.mjs: a settings guard, install and test while logged out, one Cloudflare login with seven scopes, unpaid live checks, and a logout that always runs. See docs/DEPLOY.md.Account-specific values (Drive folder IDs, the
workers.devsubdomain, machine paths) live in a settings file on the deploy machine, not here. Secrets live only in Cloudflare.
Running the tests
Node.js 22 or later:
npm ci
npm test
npm run test:runtime
python test/browser/consent_browser.py # needs Playwright with ChromiumNothing here needs a Cloudflare login, a key or a network service.
Licence
MIT. Security reports: see SECURITY.md.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
This server cannot be deployed
Maintenance
Related MCP Connectors
Give Claude only the Google Drive files you choose. Every action logged.
Paid remote MCP for Claude Code skill update gate MCP, structured receipts, audit logs, and reviewer
AI Visibility and Content Intelligence tools for Claude and MCP-compatible agents.
Paid remote MCP for LLM security scans, jailbreak checks, analytics, checkout, and readiness.
Related MCP Servers
- AlicenseAqualityCmaintenanceProvides real-time oversight for Claude Code by connecting it to an OpenClaw agent that reviews, approves, blocks, or modifies actions before they execute.3MIT
- AlicenseNot gradedqualityAmaintenanceEnables Claude.ai and Claude Desktop clients to securely connect via OAuth 2.1 to self-hosted OpenClaw gateways, offering chat, health checks, async task management, and multi-instance routing.406 npmMIT
- AlicenseNot gradedqualityAmaintenanceEnables Claude.ai and Claude Desktop to securely connect to self-hosted OpenClaw gateways through OAuth 2.1, supporting chat, health checks, and asynchronous task management across multiple instances.406 npmMIT
- AlicenseNot gradedqualityCmaintenanceLets Claude Code sign in through Microsoft Entra ID and call infrastructure tools, while privileged actions stay blocked until a human approves that exact request on a web page. Grants are bound to the user, tool, and arguments, expire automatically, and can be revoked, so a valid token alone never grants standing access to production changes.MIT