MCP server
Moonborn's Model Context Protocol server exposes personas as resources and chat as a tool. Connect Claude / Cursor / agent frameworks with a single config block.
The Model Context Protocol (MCP) is Anthropic's open standard for
LLM-tool interop. Moonborn ships an MCP server at
https://api.moonborn.co/v1/mcp — any MCP-compatible client can list
your workspace's personas as resources and invoke chat sessions
as a tool.
Wiring
The server URL + bearer auth is the whole connection. Per-client config varies; the contract doesn't:
{
"mcpServers": {
"moonborn": {
"transport": "https",
"url": "https://api.moonborn.co/v1/mcp",
"headers": {
"Authorization": "Bearer ${MOONBORN_API_KEY}"
}
}
}
}Confirmed clients: Claude for VS Code, Cursor, JetBrains MCP plugin, LangChain MCP integration, LlamaIndex.
What's exposed
Resources
moonborn:personas/{id}— read-only access to the persona's full four-layer document, voice fingerprint ID, and audit verdict.moonborn:personas— listing resource that enumerates the workspace's personas.
Tools
moonborn.chat— open a session against a persona, send a message, return the reply + drift envelope.
Scopes required
MCP authentication is the same bearer system as the REST API. Minimum scopes:
read:personasto list and read resources.write:chatto invoke the chat tool.
Issue a scoped API key per MCP client. Don't share a master key with your IDE — least privilege wins.
What's not on MCP
- No write access to personas (no create, no update, no delete).
- No config access.
- No audit log access.
- No webhook management.
The MCP server is intentionally narrow: it's the IDE/agent-friendly read + chat surface. For full read/write, hit the REST API.
Telemetry
Every MCP call lands in the standard audit log. The session ID
returned by moonborn.chat is the same session ID you'd see hitting
/v1/chat/sessions/{id} over REST.
Tier
Team and up.
Honest scope
MCP is a protocol bridge, not a separate product. It's the same runtime, the same auth, the same audit trail — just packaged for MCP-aware tooling.
Next
- Setup walkthrough: MCP server integration tutorial.
- Combined drop-in framing: Drop in (OpenAI + MCP) use case.
- MCP API reference.