Skip to main content
Your code talks to Lawtte through the REST API at https://www.lawtte.ai/api/v1, and Lawtte can push each call to you with signed webhooks.
Using Claude or ChatGPT? You don’t need any of this — add the connector and sign in. See connect Claude.

​
1. Create a connector key

Open your Studio agent’s Settings → AI connectors → Connector keys, name the key, and choose its permission:
  • Read call history — read calls, transcripts, saved agents and the Do Not Call list, and check a call before placing it.
  • Read history and place calls — all of that, plus placing calls and editing saved agents and the Do Not Call list.
The key is shown once. Keep it server-side: a call-enabled key dials real people.

​
2. Read your calls

You get back { "calls": [...] } — the calls your receptionist answered and the ones you placed, newest first. A 401 means the key is wrong, revoked, or expired.

​
3. Check a call before dialing

POST /calls/check runs every rule a real call would — consent, calling hours, Do Not Call, daily limits — and dials nothing.
{ "allowed": true, ... } means POST /calls with the same body would dial. Otherwise reason says why not.

​
4. Place the call

This dials a real person. Test with your own number first, and read limits and safety before calling lists.
Send the same body to POST /api/v1/calls. You get 201 with a call_id, or 422 with the reason Lawtte refused.

​
5. Get the result

Poll GET /api/v1/calls/{call_id} for the outcome, summary, and duration, and GET /api/v1/calls/{call_id}/transcript for what was said. Or add an endpoint in Settings → Webhooks and Lawtte posts call_ended and call_analyzed events to you. See call webhooks.

​
Next steps

API reference

Every endpoint, with request and response fields and a live playground.

Saved outbound agents

Set a call up once, run it many times with agent_id.