Skip to main content

​
MCP Integration

New to MCP? The Model Context Protocol is an open standard that lets AI assistants (Claude, Cursor, VS Code Copilot, and many others) call external tools and services. You add one server entry to your AI tool’s config file, and the tool discovers that server’s capabilities automatically. No SDK, no API client code, nothing to build.
The Model Context Protocol (MCP) lets AI tools connect to external services. SuperDocs exposes an MCP server with 38 tools and 4 user-invocable workflow prompts for document editing, session management, attachments, multi-document and durable Files, revert, and more — all on a single endpoint at https://api.superdocs.app/mcp/. When connected, SuperDocs tools appear natively in your AI tool — no API client code needed.
Running as an autonomous AI agent? You can create your own account and API key in a single request (no web form, no human), then check your quota anytime with the get_account_status tool. When you near your monthly cap, hand the account to a human with a one-time takeover code: they open an emailed link, enter the code, and adopt the account in place so your work and API key keep working. Full flow (signup, quota, handoff, adopt) is on the For AI Agents page.

​
Where to find SuperDocs

You can install SuperDocs from any of these public channels (most users only need one): Two conveniences tie these channels together:
  • The plugin and the skill both include a “Setup check” snippet that tells your AI agent to add the MCP server itself (once per session, not per tool call) if it’s not already connected. Installing the skill alone is enough to get a working integration on any agent that can run shell commands.
  • The reverse also works: connecting the MCP server alone surfaces a recommendation to load the skill, via the MCP instructions field returned during the protocol handshake.

​
Prerequisites

​
Connection details

Both /mcp and /mcp/ are accepted; the trailing-slash form is the canonical spelling used throughout these docs.
Keep your key out of dotfiles and repos. MCP config files (.mcp.json, .cursor/mcp.json, .vscode/mcp.json) are easy to commit by accident. Where your client supports environment-variable references in its MCP config (for example ${env:SUPERDOCS_API_KEY} in VS Code’s mcp.json), prefer that over pasting the raw sk_ key inline — and never commit a config containing a real key. If a key does leak, revoke it in API Keys and create a new one.

​
Supported clients

Claude Code

One install command — plugin bundles tools + prompts + skill

Claude Desktop

Add to claude_desktop_config.json

Cursor & VS Code

Add to .cursor/mcp.json or .vscode/mcp.json
These are just the most popular clients. Any application with an MCP client can connect to SuperDocs — including your own custom AI agents, internal business tools, or startup products. If it speaks MCP, it works with SuperDocs. Just point it at the endpoint and auth details above.

​
Quick test

After connecting, ask your AI tool:
“Use the SuperDocs health tool to check the API status”
If the connection is working, you’ll get back {"status": "healthy"}.
The MCP health tool runs over your authenticated /mcp/ connection, like every MCP tool, so it doubles as a check that your API key is set up correctly. A 401 here means the key is missing or misconfigured, not that the API is down. If you want a zero-auth status probe, use the REST endpoint instead: curl https://api.superdocs.app/health requires no authentication.

​
Troubleshooting

Tools panel says “Loading tools” forever — restart your AI tool. MCP servers load at startup, so a fresh session is required after adding the server. Connection fails or times out — verify your API key works:
A 200 response (with a JSON list of your sessions, possibly empty) confirms the key is valid. A 401 means the key is wrong, revoked, or the Authorization header didn’t reach the server.