Dialfetch MCP — give your AI agent a phone ☎️

Connect Dialfetch to Claude, Cursor, VS Code, Windsurf, Cline, the Claude Agent SDK — or any Model Context Protocol client — and your agent can place real phone calls. It navigates IVR phone trees, waits on hold, talks to humans (including after a transfer), and hands back structured answers, a transcript, and a recording.

Give it a phone number and a plain-language goal. Goal in, result out.

"Call the pharmacy at +1 305-555-0148 and tell me what time it closes today." → { answers: { closing_time: { value: "9:00 PM", quote: "We close at nine tonight." } } }

🔌 Endpoint: https://app.dialfetch.com/api/mcp (Streamable HTTP + SSE)
🔑 Auth: API key (Authorization: Bearer df_…) or one-click OAuth for custom connectors
📚 Full docs: https://app.dialfetch.com/docs/mcp
🌐 Industry-agnostic: retail & pharmacy hours, flight/travel status, order & claim status, license & insurance verification, government & utility lines

Quick connect

Claude Desktop

Add to claude_desktop_config.json (Settings → Developer → Edit config):

{
  "mcpServers": {
    "dialfetch": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://app.dialfetch.com/api/mcp",
               "--header", "Authorization: Bearer df_YOUR_KEY"]
    }
  }
}

Claude.ai (web) — custom connector

Settings → Connectors → Add custom connector → paste https://app.dialfetch.com/api/mcp and finish the one-click OAuth sign-in. No API key to paste.

Cursor

~/.cursor/mcp.json:

{
  "mcpServers": {
    "dialfetch": {
      "url": "https://app.dialfetch.com/api/mcp",
      "headers": { "Authorization": "Bearer df_YOUR_KEY" }
    }
  }
}

VS Code (Copilot agent mode)

.vscode/mcp.json:

{
  "servers": {
    "dialfetch": {
      "type": "http",
      "url": "https://app.dialfetch.com/api/mcp",
      "headers": { "Authorization": "Bearer df_YOUR_KEY" }
    }
  }
}

More configs (Windsurf, Cline, Agent SDK) are in configs/.

Tools

Tool	Inputs	Does
`place_call`	`phone, goal, extract?, context?, disclose_ai?, from_number_id?, max_duration_minutes?, webhook_url?`	Start a goal-call; returns `{ call_id, status }` (async)
`get_call`	`call_id`	Status, outcome, answers, notes, transcript, recording URL
`wait_for_call`	`call_id, timeout_seconds?`	Block (with progress) until the call is terminal
`cancel_call`	`call_id`	Cancel a queued / in-progress call
`list_calls`	`status?, limit?, cursor?`	Recent calls, newest first
`get_call_recording`	`call_id`	Short-lived presigned audio URL
`list_phone_numbers`	—	Caller-ID numbers you own
`get_pricing`	—	Per-call / per-number pricing
`get_credit_balance`	—	Prepaid credit balance

Resource: dialfetch://calls/{id} · Prompts: check_store_hours, verify_license, get_order_status

How calls work

Calls are asynchronous (typically 2–10 minutes; longer with holds or transfers):

place_call({ phone, goal, extract }) → { call_id, status: "queued" }
Poll get_call({ call_id }) (or use wait_for_call) until status is completed / failed / canceled
Read answers, transcript, recording_url

Recipes

Safety

Calls disclose they are an AI by default. The agent will not read out or collect sensitive personal data (SSNs, passwords, full card numbers). Goal-screening, credit holds, and rate limits apply to every call and can't be bypassed by a client.

License

MIT — see LICENSE.

com.dialfetch/dialfetch

Links

README