Skip to content

Claude Code

Claude Code adds MCP servers from the terminal with claude mcp add. This is the path to take when you want the agent writing integration code with the docs right there in its tools — see Let the AI do the coding.

No clone, no build. Add the deployed Workers over HTTP:

Terminal window
# Rails MCP — operate the lifecycle (token-gated; token + slot in the URL)
claude mcp add --transport http partifact-rails \
"https://partifact-rails-mcp.thanhvuttv.workers.dev/mcp/<MCP_ACCESS_TOKEN>/<your-slot>"
# Docs MCP — read the docs as tools (public, no token)
claude mcp add --transport http partifact-docs \
"https://partifact-docs-mcp.thanhvuttv.workers.dev/mcp"

Replace <MCP_ACCESS_TOKEN> with the token from your invite and <your-slot> with any name ([a-z0-9_-]). Prefer to keep the token out of the URL? Put it in a header instead and keep the slot in the path:

Terminal window
claude mcp add --transport http partifact-rails \
--header "Authorization: Bearer <MCP_ACCESS_TOKEN>" \
"https://partifact-rails-mcp.thanhvuttv.workers.dev/mcp/<your-slot>"

claude mcp add defaults to local scope (this project, just you). Two others matter:

  • --scope user — available across all your projects.
  • --scope project — written to a .mcp.json in the repo, so it’s shared with everyone who clones it (and the agent gets the tools automatically). Don’t commit a real token this way.

A project .mcp.json looks like:

{
"mcpServers": {
"partifact-rails": {
"type": "http",
"url": "https://partifact-rails-mcp.thanhvuttv.workers.dev/mcp/<MCP_ACCESS_TOKEN>/<your-slot>"
},
"partifact-docs": {
"type": "http",
"url": "https://partifact-docs-mcp.thanhvuttv.workers.dev/mcp"
}
}
}
Terminal window
claude mcp list # shows each server + a connection check
claude mcp get partifact-rails

Inside a Claude Code session, /mcp lists connected servers and their tools. Then hand it the work — drive the Corolla job with the Rails MCP, or point it at the API and have it build an integration.