Telnyx Meeting API Now Available in Beta

Telnyx Meeting API is now available in beta. Point it at a Google Meet, Zoom, Microsoft Teams, or Webex meeting and a programmable AI participant joins the room: it transcribes live, can speak and post to chat, returns structured post-meeting artifacts, and can be driven end to end by a Telnyx AI Assistant, all through one API key.

What's new

• Meeting API: A REST surface on /v2/meeting_sessions to create or schedule a session for an existing meeting, follow the live transcript, drive the bot with actions/speak, actions/stop_speaking, and actions/send_chat, and collect recordings after the meeting ends.
• AI Assistant integration: Attach any Telnyx AI Assistant to a session with its ID, and it listens to the room, reasons with your configured model and tools, and speaks back in real time in its own voice. Dynamic variables personalize the assistant per meeting, and full_duplex audio lets participants interrupt it.
• Six artifact types: Request summary, action_items, decisions, topics, open_questions, or custom, where custom answers any prompt you supply, grounded in the transcript alone. summarize_on_end auto-generates a summary when the meeting ends. Artifacts run on Telnyx Inference.
• Digital avatar support: Bring your own Anam key and the bot renders as a lip-synced avatar on camera, alone or combined with an attached assistant. A custom still image on the bot's camera tile is also supported.
• Arrival announcements: speak_on_enter and chat_on_enter greet the meeting by voice or chat the moment the bot is admitted.
• Signed webhooks: Set webhook_url on session create to receive signed lifecycle events: session status changes, transcript completion, recording availability, and artifact completion.
• MCP server: Connect agents through 19 dedicated Meeting tools over Streamable HTTP at /v2/meeting_bot/mcp: joining meetings, reading transcripts and events, in-meeting actions, artifacts, and calendar management.
• Calendar auto-join: Connect a Google Calendar once and the bot automatically joins meetings with a link, with connection-level policy and per-meeting overrides. Included at no charge.
• Recording control: Sessions record automatically. Recordings are retrievable with short-lived download URLs, and recording media can be deleted without touching the transcript or artifacts.
• Account-scoped tenancy: Every operation authenticates with a Telnyx API key and is scoped to the authenticated account. Foreign and missing IDs both return 404.

Pricing

• Standard session: $0.02 per minute, which is bot presence at $0.0126/min plus live transcription at the Deepgram Nova 3 rate ($0.0074/min). Bot speech bills per character at $0.000003. Recordings are free for the first 7 days of retention, then $0.05 per recording media hour per 30 days.
• Assistant session: An attached AI Assistant bills $0.05/min plus LLM tokens in place of transcription and speech, so only bot presence continues alongside it. Digital avatars are bring-your-own-key and bill on the provider side.

Why it matters

• Agents have been shut out of meetings. A meeting is where decisions happen, but until now an agent could not join one, hear what was said, or contribute. Meeting API gives an agent live and durable meeting context through the same API key it already uses for calls, STT, TTS, and inference.
• Granola and similar tools record meetings for humans. Meeting API is for agents. A developer can send an agent into a standup it cannot attend, stream the transcript back, raise a prompt through speech, and retain the full meeting context afterwards.
• Unlike record-only notetakers, an attached Telnyx AI Assistant does not just transcribe: it participates. It listens to the room, reasons with the model and tools you configured, and speaks when addressed, all inside the meeting.
• Meeting platforms each ship their own recorder, transcription, and bot APIs with different contracts and rate limits. Meeting API abstracts that behind one Telnyx-shaped contract, so the agent does not learn four vendor SDKs.
• One API key covers the meeting, the assistant, the speech, and the inference. No second vendor for meeting presence, no second contract for transcription.

Example use cases

• An engineering team brings its agent into a technical brainstorming session so it retains the discussion and can contribute when instructed.
• A user sends the bot to a standup they cannot attend. The bot takes the transcript, asks its questions through speech, and returns the meeting context afterwards.
• A support team records every escalation call through one pipeline, then generates summaries, decisions, and action items through Telnyx Inference.
• A sales team pipes meeting transcripts into their CRM workflow, with structured artifacts instead of manual note-taking.
• A developer building a meeting-aware agent attaches an AI Assistant, gives it tools, and lets it act inside the meeting on their behalf.

Getting started

1. Get a Telnyx API key from Mission Control.
2. Create a session with POST https://api.telnyx.com/v2/meeting_sessions, passing meeting_url and bot_name. Optional: webhook_url, summarize_on_end, chat_on_enter, speak_on_enter, camera_image, and an assistant to automate the conversation.
3. Poll GET /v2/meeting_sessions/{id} until the status is active. If the platform requests admission, admit the bot in the meeting UI.
4. Follow the live transcript with GET /v2/meeting_sessions/{id}/transcript or the WebSocket stream, or let an attached assistant handle the conversation on its own.
5. Drive the bot manually with actions/speak, actions/stop_speaking, and actions/send_chat.
6. After the meeting ends, request artifacts (summary, action_items, decisions, topics, open_questions, or custom), fetch recordings with GET /v2/meeting_sessions/{id}/recordings, or delete recording media when you only need the transcript.

Learn more in the Telnyx API reference or the Telnyx developer docs.