Connect a client
Get your MCP URL and wire it into Claude or ChatGPT
Tool reference
Every tool, its inputs and what it returns
What the assistant can do
Architecture
The MCP server runs inside the Ringee backend. There is nothing extra to deploy. Every tool call resolves to the same domain services the dashboard uses, so permissions, credit rules and workspace scoping behave identically no matter where the request comes from.Two endpoints, two auth models
Workspace scoping
Everything an assistant reads or writes is scoped to the active workspace — your personal account or one organization. The assistant can move between them withswitch_workspace, with no re-authentication. See Workspaces.
Safety model
Tools are annotated with the MCP trust hints (readOnlyHint, destructiveHint, idempotentHint, openWorldHint) so clients can warn before something consequential happens.
- Destructive tools include contact deletion, session revocation, campaign-lead removal and DNC removal. Their confirmation rules depend on the target.
- Sensitive tools include lead reveal, call-session creation and
start_ai_voice_agent_call. start_ai_voice_agent_callstarts a real, billed call.reveal_leaduses the connected enrichment provider’s allowance instead of Ringee calling credits; lead search itself is free.
Beyond the raw protocol
ringee CLI
Drive the same tools from a terminal or a shell-capable agent
Claude skills
/ringee commands in Claude Code and claude.aiChatGPT app
Visual Ringee cards inside ChatGPT
Next steps
Connect your first client
Get your MCP URL from the dashboard and paste it into Claude

