Docs · MCP

Operon MCP server

A Model Context Protocol server that lets your agent search and fetch operons (versioned AI skills) from your team's catalog, and record runs for governance.

Endpoint

https://mcp.withoperon.com/mcp

Transport

streamable-http

Auth

OAuth2 · Supabase

Tools

The server exposes three tools. Read-only access to the catalog, plus a write call for run history.

list_operons

List operons visible to the calling workspace member.

inputs: workspace (optional slug), query (optional substring).
returns: Array of { slug, semver, visibility, lastUpdatedAt }.

get_operon

Fetch a single operon's SKILL.md and version metadata so the agent can load it as context.

inputs: slug (required), version (optional, defaults to latest).
returns: { slug, semver, markdown, scope, permissions, createdAt }.

record_run

Record an invocation against an operon version for run history and cost telemetry.

inputs: slug, version, status (ok | err | skip), latencyMs, costUsd, notes (optional).
returns: { runId }.

Hand off to your agent

Paste this into Claude Code, Cursor, or Codex and the agent will wire up the MCP server in the right config file for you. Works even if it has never seen Operon before.

agent prompt

Add the Operon MCP server to my agent configuration.

Endpoint:       https://mcp.withoperon.com/mcp
Transport:      streamable-http
Authentication: OAuth2 via Supabase (the server's discovery doc is at
                https://withoperon.com/.well-known/oauth-protected-resource).

Pick the right config for the client I'm running in:

  Claude Code:
    Run: claude mcp add operon --transport http https://mcp.withoperon.com/mcp

  Cursor:
    Edit ~/.cursor/mcp.json and add:
      { "mcpServers": { "operon": { "url": "https://mcp.withoperon.com/mcp" } } }

  Codex:
    Edit ~/.codex/config.json and add:
      { "mcpServers": { "operon": { "url": "https://mcp.withoperon.com/mcp" } } }

After it's wired up, the server exposes three tools:
  - list_operons : list operons visible to my workspace
  - get_operon   : fetch a specific operon's SKILL.md + version metadata
  - record_run   : record an invocation so it appears in run history / cost

Docs: https://withoperon.com/docs/mcp

Or set it up by hand

Once configured, your agent gets list/get tools for every operon you can see, plus a record-run tool so every invocation is auditable.

Claude Code

                claude mcp add operon --transport http https://mcp.withoperon.com/mcp

Cursor

                ~/.cursor/mcp.json → add { "operon": { "url": "https://mcp.withoperon.com/mcp" } }

Codex

                ~/.codex/config.json → mcpServers.operon = { "url": "https://mcp.withoperon.com/mcp" }

Discovery

The server's machine‑readable card lives at /.well-known/mcp/server-card.json. Authentication discovery is at /.well-known/oauth-protected-resource.