api.agntix.ai gateway documented in the API Reference, with the same authentication and RBAC scopes. MCP is simply a second way to reach the platform — built for AI clients instead of your own code.
What you can do
Manage agents
List, create, update, and delete agents — including their model, system prompt, tools, and voice pipeline.
Drive chat sessions
Open sessions, send messages, and read transcripts on behalf of any agent.
Configure tools
Inspect and wire up API/function tools that your agents call at runtime.
Operate telephony
Provision phone numbers and trigger outbound calls and campaigns.
Inspect analytics
Pull usage, cost, and call-outcome data to summarize or report on.
Stay in scope
Tools respect the scopes on your API key — a read-only key can only read.
MCP vs. the REST API
Both reach the same platform through the same gateway. Choose based on who is making the call.| MCP server | REST API | |
|---|---|---|
| Caller | An AI client (Claude, Cursor, your agent) | Your own application code |
| Interface | Natural-language tool calls | HTTP requests you write |
| Auth | Org API key / Bearer JWT | Org API key / Bearer JWT |
| Enforcement | Same RBAC scopes & rate limits | Same RBAC scopes & rate limits |
| Best for | Letting an assistant operate Agntix for you | Deterministic, programmatic integrations |
Every MCP tool maps 1:1 to a documented REST endpoint. If you can’t do it over the API, you can’t do it over MCP — and vice versa.
How it works
- Your MCP client connects to the Agntix MCP server and discovers the available tools (
tools/list). - When the assistant decides to act, it issues a
tools/call. - The MCP server forwards the call to the
api.agntix.aigateway with your API key. - The gateway applies the same authentication, RBAC, and rate limits as a direct API call, then returns the result to the assistant.
Next steps
Connect
Server URL, transport, and authentication.
Client setup
Configure Claude Desktop, Cursor, and others.
Tools reference
Every tool and the endpoint it maps to.