@kanopi/callrail-mcp

A Model Context Protocol (MCP) server for the CallRail API v3. It exposes CallRail's call tracking, text messaging, form submissions, trackers, companies, tags and more as MCP tools so an LLM client (Claude Desktop, Claude Code, etc.) can read from and write to your CallRail account.

• Transport: stdio (local)

• Auth: CallRail API v3 key via environment variable

• Coverage: 34 tools across 11 resources (full read + write)

Prerequisites

• Node.js 18+ (uses the global fetch)

• A CallRail API v3 key — in CallRail go to Settings → Integrations → API Keys → Create New API v3 Key. Keys are user-scoped; the server can only see what that user can.

Related MCP server: MCP API Server

Usage

No install needed — run the published package directly with npx:

npx -y @kanopi/callrail-mcp

npx fetches @kanopi/callrail-mcp and runs its callrail-mcp binary over stdio. This is what you point your MCP client at (see Connecting to a client).

Build from source (for development)

git clone <repo-url> && cd call-rail-mcp
npm install
npm run build

This compiles TypeScript to dist/. The entrypoint is dist/index.js.

Configuration

The server reads credentials from environment variables:

Copy .env.example to .env for local reference (the server reads from the process environment; your MCP client passes these in — see below).

Connecting to a client

Claude Code

claude mcp add callrail \
  --env CALLRAIL_API_KEY=your_api_key_here \
  --env CALLRAIL_ACCOUNT_ID=your_account_id \
  -- npx -y @kanopi/callrail-mcp

Claude Desktop / other clients (JSON config)

Add to the mcpServers block of the client's config (e.g. claude_desktop_config.json):

{
  "mcpServers": {
    "callrail": {
      "command": "npx",
      "args": ["-y", "@kanopi/callrail-mcp"],
      "env": {
        "CALLRAIL_API_KEY": "your_api_key_here",
        "CALLRAIL_ACCOUNT_ID": "your_account_id"
      }
    }
  }
}

To run a local checkout instead of the published package, point at the built entrypoint — "command": "node", "args": ["/absolute/path/to/call-rail-mcp/dist/index.js"] — or the TypeScript source via tsx: "command": "npx", "args": ["tsx", "/abs/path/src/index.ts"].

Tools

Write tools are prefixed with WRITE: in their descriptions and carry MCP annotations (readOnlyHint/destructiveHint) so clients can prompt for confirmation. Most tools accept an optional account_id that falls back to CALLRAIL_ACCOUNT_ID. List tools share common params: page, per_page (max 250), sort, order, search, date_range/start_date/ end_date, and fields for optional extra fields.

✏️ = write (create/update) · 🗑️ = destructive (delete)

Notes & caveats

• Premium data: transcription, sentiment, call summaries and highlights require a CallRail Premium Conversation Intelligence subscription. Request them via the fields argument on get_call / list_calls.

• SMS compliance: outbound texts must include opt-out instructions; CallRail enforces this.

• Rate limits: ~1,000 requests/hour by default. The server surfaces HTTP 429 (with Retry-After when provided) as a tool error rather than crashing.

• Data retention: communication records are retained ~25 months; older date ranges error.

• Pagination is manual: pass page to walk results; the server does not auto-page.

Development

npm run dev     # run from source with tsx
npm run build   # type-check + emit dist/
npm start       # run the built server

Project layout

src/
  index.ts            entrypoint (stdio transport)
  config.ts           env loading + validation
  client.ts           CallRail HTTP client (auth, query, errors, account resolution)
  types.ts            shared types + CallRailApiError
  tools/
    index.ts          registers all tools
    shared.ts         common zod params, response/error helpers, annotations
    <resource>.ts     one file per CallRail resource

License

MIT