This server pulls in the curated collection of GitHub Copilot agents, prompts, instructions, skills, and workflows from the awesome-copilot repository and exposes them through MCP. You get search across all these resources, on-demand content retrieval, and a refresh tool to pull the latest updates from GitHub. It bundles a metadata snapshot directly in the npm package so it works immediately without hitting the network, then lets you hot-reload from the live repo when you need current data. Useful when you want to browse and integrate community Copilot patterns without leaving your MCP client. Also runs as an HTTP server with OpenAPI docs if you need remote access or want to test the endpoints directly.
A Model Context Protocol (MCP) server that provides access to awesome-copilot agents and collection resources.
Click the badge to install directly in VS Code (will prompt to open VS Code):
Note: Requires GitHub Copilot extension in VS Code. Click the badge on GitHub to trigger the installation prompt.
Add to your MCP Client configuration (e.g., Claude Desktop or VS Code):
{
"mcpServers": {
"awesome-copilot": {
"command": "npx",
"args": ["-y", "awesome-copilot-mcp", "start"]
}
}
}
This ensures you always run the latest version.
Standard stdio usage (default). See configuration above.
Run locally for remote access or OpenAPI testing:
npx -y awesome-copilot-mcp start-http --port 8080 --host 0.0.0.0
Available endpoints:
GET /health Health checkGET /metadata Return metadata indexGET /metadata/stream SSE streaming output of metadataGET /search?q=kw SearchGET /openapi.json API DocumentationThis server now indexes all resources published in the official github/awesome-copilot repository:
plugins/external.json)MCP Inspector is the recommended way to test and debug:
# Debug via Stdio (Recommended)
npx -y awesome-copilot-mcp debug --no-build
# Debug via HTTP
npm run inspect:http
This server uses a Bundled + In-Memory architecture for maximum reliability:
metadata.json bundled directly within the npm package. Zero external dependencies.refresh_metadata fetches the latest data from GitHub and stores it in memory for the current session.~/.cache) is used, preventing permission issues in restricted environments (e.g., macOS App Sandbox).| Variable | Description |
|---|---|
ACP_METADATA_URL | Optional. URL to a hosted metadata.json (e.g. https://yourname.github.io/repo/metadata.json). Overrides GitHub raw fetch. |
ACP_REPOS_JSON | Custom repository configuration JSON. |
Automated: GitHub Actions runs daily to fetch the latest metadata from github/awesome-copilot and publishes a new npm version if changes are detected.
Internal: The metadata file is lightweight (~170 KB) containing only index info. Actual content and download targets for agents, prompts, instructions, skills, collections, plugins, hooks, and workflows are resolved on-demand via the exposed tools and resources.
# Install dependencies
npm install
# Build
npm run build
# Run tests
npm test
MIT
io.github.ericm1018/skillfm-llm-cost-optimizer-openai-anthropic-usage
io.github.mikerawsonnz/llm-orchestration-agent
io.github.mikerawsonnz/authenticated-llm-agent
labforgedev/copilot-memory-mcp
csoai-org/agent-prompt-injection-firewall-mcp
io.github.mikerawsonnz/authenticated-multi-llm-agent