Golem CLI comes with http streamable mcp server, enables you to interact with all golem commands using any agent such as Claude Caude or Openai Codex.

All relevant Golem CLI commands are available as tools, refer to this to see them in action: https://youtu.be/t5dnCSYQg_0

Golem MCP server will provide all existing manifests (yaml or yml) in the current working directory (where the mcp launches) as resources, you can then read such manifest from such llm agent using just single prompt like: what's inside @suchmanifest (drop down menu will appear, just when type @), refer to this video: https://youtu.be/95BXexeZjj4

Try to build the package with features enabled

cargo build -p golem-cli --release --features "server-commands"

then start the binary

./target/release/golem-cli --serve --serve-port 1232

Golem also comes with basic http streamable mcp client, enables you to interact with the Golem MCP server if you want standlaone client or depends of such use case.

Try to make the ports identical (edit tests/mcp-client/src/main.rs to match the server port), then build the package

cargo build -p golem-mcp-client --release

then start the binary

./target/release/golem-mcp-client

There is a video for the e2e mcp server/client testing: https://youtu.be/ONUJ6BOyHDI