AI Configuration

You can drive Prelude TE from any AI tool that speaks the Model Context Protocol — Claude Code, Cursor, Codex, and others. This section describes the two pieces that make that work.

What you get

With both pieces installed, an AI assistant sitting next to your network can:

• Read the topology — list domains, walk nodes and links, look at Flex-Algo views.
• Manage BGP-LS peers — add a new peer, enable or disable an existing one, watch its session state.
• Manage outputs — declare a NATS output, check its connection state, edit credentials.
• Sanity-check connectivity to a Prelude TE instance with a ping tool.

All of this happens through the same REST API a human would use, behind the assistant's chat.

The two pieces

• Prelude MCP server — a separate, stateless service that exposes Prelude TE's REST API over MCP. You point your AI tool's MCP configuration at this server.
• Companion skill (roadmap) — a reusable skill bundle that teaches the assistant how to drive the tools the server exposes. Not yet available; in the meantime the assistant uses the MCP tools directly, without a Prelude-specific playbook.

For the exhaustive tool list, see Available tools.