Cloudpen Speaks MCP: Connect Claude, Cursor and Your

AI tools are only as useful as the context they can see. The Model Context Protocol (MCP) is an open standard for giving them that context, and Cloudpen supports it in both directions. Both live under the MCP tab in the editor sidebar, are scoped per project, and are completely independent of each other.

Direction 1: Quill as an MCP client (Data Sources)

Connect Quill to a live backend such as Supabase, Postgres, Neon, PlanetScale, GitHub, or any custom MCP endpoint. Quill can then look at your real schema and tables before it writes code, instead of guessing or asking you to paste them in.

• Read-only: Quill can look, not write. No migrations, no data changes, nothing destructive
• Opt-in auto-fetch: you decide per connection what Quill can see and when
• Test first: verify any connection from the panel before relying on it

On the Pro plan you get 50 MCP calls per day. Team and Enterprise are unlimited. Find it under MCP → Data Sources.

Direction 2: Cloudpen as an MCP server

The other way around, you can let external AI tools connect to a Cloudpen project. Claude Desktop, Claude Code, Cursor, or any MCP-compatible client can read your files, check deployment status and see your git history from its own interface.

Security was the starting point for this feature:

• Per-project tokens: each token is scoped to exactly one project, and revoking one never affects another
• Read by default: file listing and file contents are on for any active token
• Writes are opt-in: the AI client must call a confirm step before any file is actually written
• Email confirmation: creating a write-scoped token requires a one-time code sent to your email
• Hard 90-day expiry on write-scoped tokens, whatever you configure
• Activity log: every tool call is recorded with what was called, when and from which IP

Connect an AI client in five steps

1. Open a project and go to MCP → MCP Server
2. Create a token with a label (for example "Claude Desktop") and choose read-only or read and write
3. If you enabled write access, confirm with the code emailed to you
4. Copy the token immediately, because it is shown only once
5. Use it as the bearer token for https://mcp.cloudpen.dev in your client's MCP configuration

The MCP Server is included on Pro (up to 5 active tokens), Team and Enterprise, as well as on eligible AppSumo tiers. On Free, the panel is visible but creating a token prompts you to upgrade.

Closing the edit and test loop

MCP also powers Cloudpen's Chrome Extension Testing. A connected AI assistant can read console errors from your extension and trigger a reload itself, so it can edit code, test it and fix it without you relaying error messages back and forth.

Give your AI tools a safe window into your project, and let them do more of the work.