GitHub

The Model Context Protocol (MCP) is an open standard that enables AI models to interact with external tools and data. Compeer provides a built-in MCP server that allows you to expose your captures and stores to AI agents.

Spawning a Server

You can spawn a local MCP server using the compeer mcp command. This server communicates over stdio, making it compatible with most MCP hosts.

compeer mcp [workspace] [store]

[workspace]: The name of the workspace to expose.
[store]: (Optional) Limit the server’s context to a specific store.

Available Tools

The Compeer MCP server provides the following tools to the AI model:

`search`

Searches for captures within the configured workspace and store.

query (string): The search query to find relevant context.

Common Harnesses

You can connect the Compeer MCP server to various AI assistants and IDEs.

OpenCode: A model-agnostic CLI for agentic workflows.
Claude Code: Anthropic’s terminal-based assistant.
OpenAI: Use with the ChatGPT Desktop app via Developer Mode.
GitHub Copilot: Integrate Compeer context into your IDE.

Configuration Examples

OpenCode

To use Compeer with OpenCode, add the following to your .opencode/mcp.json. This example assume you have a workspace named “product” and a store named “design-decisisons”:

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "compeer-product-design-decisisons": {
      "command": "compeer",
      "args": ["mcp", "product", "design-decisisons"],
      "enabled": true,
    },
  },
}