# For agents and automation — MCP

> Operate the Poneva platform from an AI agent over the Model Context Protocol.

The Poneva platform is agent-operable over the Model Context Protocol: an AI agent can
drive the same platform actions a human drives in the dashboard, authenticated as your
account and bounded by the same permissions.

## Connecting

The MCP server runs over Streamable HTTP at `https://poneva.com/mcp`, authorized via OAuth
2.0 Device Authorization (RFC 8628). Discovery surfaces for agents:

- **Server card** — [https://poneva.com/.well-known/mcp.json](https://poneva.com/.well-known/mcp.json)
  carries the endpoint, authorization endpoints, and the full tool list.
- **Agent-readable overview** — [https://poneva.com/llms.txt](https://poneva.com/llms.txt)
  describes the platform and each tool in plain text, and maps every page of this
  documentation with its one-line summary. It also points agents at the State of
  Agentic Commerce reference page and machine-readable JSON snapshot.
- **Raw markdown docs** — every page of this documentation is also served as plain
  markdown: append `.md` to its URL (this page is `/docs/agents-mcp.md`). The full
  corpus in one response: [https://poneva.com/llms-full.txt](https://poneva.com/llms-full.txt).

## Configuration through tools

For programmatic configuration, the `update_assistant` tool changes assistant settings
through the same validated path as the dashboard.