MCP Server

Frontend exposes a Model Context Protocol (MCP) server so AI agents — Claude, Cursor, or any MCP-compatible client — can work with your projects directly: read and write files, run git and database operations, manage environment variables and domains, generate media, and publish.

Endpoint

https://mcp.frontend.co/mcp

The server speaks the streamable HTTP transport. It runs on the core Frontend app, so https://www.frontend.co/mcp resolves to the same endpoint.

Authentication

The MCP server uses the same team API key as the REST API, sent as a Bearer token. Create one from Team Settings → API Keys (/teams/{handle}/api-keys). The key is team-scoped — it can reach any project the team owns.

There is no cookie or session auth. Configure your MCP client to send:

Authorization: Bearer <your-api-key>

Connecting a client

Most MCP clients accept a remote server URL and a header. Example configuration:

{
  "mcpServers": {
    "frontend": {
      "url": "https://mcp.frontend.co/mcp",
      "headers": { "Authorization": "Bearer <your-api-key>" }
    }
  }
}

To explore the server interactively, run the MCP Inspector against the same URL and header:

npx @modelcontextprotocol/inspector

How it works

No project is selected when you connect. An agent first calls the projects tool to list or search the team’s projects, then passes the chosen projectId to every other tool. Access is authorized per call against the key’s team.

A quick tools/list check with curl:

curl -X POST https://mcp.frontend.co/mcp \
  -H "Authorization: Bearer <your-api-key>" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}'

Tools

Tool	Description
`projects`	List or search the team’s projects; returns ids to use with the other tools
`readFiles`	Read specific project files by path
`writeFile`	Create or update a file (validated and formatted)
`editFile`	Replace exact lines in a file
`bash`	Run a shell command over the project files (offline: no network)
`git`	`log`, `commit`, `branch`, `merge`, `restore`, `push`, `pull`, `pr`
`database`	Run SQL and inspect the schema of the Frontend Cloud database
`env`	List, set, and delete environment variables
`domains`	List, add, verify, and remove custom domains
`publish`	Build and deploy the project to production
`media`	List, delete, and generate media library assets

Every project-scoped tool takes a projectId argument (from the projects tool) and an optional branch.

MCP vs REST

The MCP server is optimized for AI agents (a bash shell, action-parameterized tools). For scripts and backend integrations, prefer the typed REST API and SDK, which cover the same capabilities as granular, individually-addressable endpoints.