Public API
Server to server. Sync CRM records, open a secure dialer and run AI Voice
Agents.
Dialer SDK
Browser. Embed a real dialer inside your own web app.
MCP
AI agents. Let Claude or ChatGPT read and act on Ringee data.
CLI
Terminal and scripts. The same capabilities as MCP, in a shell.
AI Voice Agents
Autonomous calling. Hold a real conversation and return a structured result.
Pick by what you are building
Side by side
Human-operated calls still run over WebRTC in a browser or mobile app. AI
Voice Agents are the deliberate exception: Ringee starts those calls
server-side, the configured agent holds the conversation, and API, MCP or CLI
can read the asynchronous result.
A typical CRM integration
Most teams combine two surfaces:1
Sync your records
Send
contact.upserted so Ringee knows who is callable, and keep externalId
stable so click-to-call and the Dialer SDK can resolve the same contact. For AI
Voice Agent webhook correlation, also send it as metadata.external_id when you
start the call. See Inbound events.2
Let agents call
Embed the Dialer SDK for an in-product experience, or use click-to-call if you want the fastest possible path.
3
Receive the activity
Subscribe to
call.completed, call.outcome.updated and recording.ready and write them into your CRM. Signed, retried, idempotent. See Outbound webhooks.Credentials at a glance
Create
cik_live_…, whsec_… and pk_live_… credentials from Integrations → Custom Integrations. The MCP URL comes from GET /api/mcp/connection-info.
Self-hosting
Every surface works against a self-hosted deployment. Replacehttps://api.ringee.io with your own backend URL:
See the Configuration reference and AI Voice Agents self-hosting.

