Skip to main content

Telnyx AI Connector

The Telnyx AI Connector is a remote MCP server. Any MCP-compatible AI assistant can use it, such as Claude and Codex. Without signing in, it lets the assistant search the public Telnyx API reference and generate code. After you sign in with your Telnyx account, it can also read your own call status, call events, and recording metadata. It is read-only: it cannot place a call, send a message, buy a number, or change anything on your account.

Which Telnyx MCP endpoint to use

Telnyx exposes more than one MCP endpoint, so pick the one that matches what you need:
  • https://api.telnyx.com/v2/ai/mcp is this connector. It is read-only: anonymous API reference search, plus authenticated read access to your own call data over OAuth.
  • https://api.telnyx.com/v2/mcp is the Telnyx Remote MCP server. It executes Telnyx API actions and authenticates with an API key.
  • https://developers.telnyx.com/mcp is the documentation search server for these docs.

Add the connector

Claude.ai, Claude Desktop, and Claude mobile

Open Settings > Connectors > Add custom connector, enter the server URL below, and leave the advanced OAuth fields empty. Telnyx registers the client automatically. The documentation tools work right away with no sign-in. When you ask for account data, Claude shows a Connect card; approve the Telnyx sign-in and Claude retries your request.

Claude Code

Searching endpoints and reading schemas work straight away. To read your own account, run /mcp, select telnyx, and choose authenticate. A browser opens for the Telnyx sign-in.

Codex CLI

Add the server to ~/.codex/config.toml:
The documentation tools work without signing in. To read your account, also run codex mcp login telnyx.

Other MCP clients

Any MCP client that supports remote servers over Streamable HTTP can use the connector. Point it at the server URL, and sign in when you want to read account data.

What you can ask

Searching the API reference needs no sign-in. Reading your account does.
  • “Which Telnyx endpoint sends an SMS, and what does it take?”
  • “Show me the schema for creating a call.”
  • “What happened during call session <id>?” (sign-in required)
  • “Was that call recorded, and how long was it?” (sign-in required)
For the account tools you provide a call, leg, or session identifier. These appear in Telnyx call webhooks, in Call Control API responses, and in results the connector itself returns.

Tools

There are five read-only tools. There is no general-purpose “call any endpoint” tool, and documenting an endpoint never makes it executable here.
  • list_api_endpoints (no sign-in): search the public Telnyx API reference, which covers 1,336 documented operations.
  • get_api_endpoint_schema (no sign-in): return the full schema for one endpoint.
  • get_call_status (sign-in): read the status of one call.
  • list_call_events (sign-in): read the event timeline for a call leg or session.
  • search_recordings (sign-in): read recording metadata. Recording media, URLs, and transcripts are never returned.

Authentication and privacy

  • Documentation is anonymous. Searching the API reference and reading schemas use no account and touch no account data.
  • Account access is per-user and short-lived. Signing in authorizes the connector for your account only. Each request uses a Telnyx credential that expires within minutes and is never stored, and the connector keeps no long-lived end-user credential.
  • Read-only and minimal. Recording searches return metadata only. Recording URLs, media URLs, and transcripts are removed before a result is built.
  • Data handling. The connector calls Telnyx’s own first-party API. No conversation data is collected. Logs contain only a request id, a route template, and a status class, with no request bodies, headers, credentials, or account identifiers.

Support