CLI reference

Entrypoint: mechaharness (mechaharness.cli.main:app)

Commands

Command	Purpose
version	Print package version
backends	List configured inference backends
families	List configured harness families
run	Run one prompt through a harness
serve	Start the FastAPI server
describe	Show metadata for a configured backend

run

Flag	Env / default	Description
PROMPT (arg)	—	User prompt
--backend / -b	MECHA_INFERENCE_BACKEND (openai)	Inference backend name
--family / -f	MECHA_HARNESS_FAMILY (tool_loop)	Harness family (pass_through, tool_loop, …)
--model / -m	MECHA_MODEL	Model id
--api-key	MECHA_API_KEY	Provider API key
--base-url	MECHA_BASE_URL	Override endpoint
--system	MECHA_SYSTEM_PROMPT	System prompt
--max-turns	8	Agent loop limit
--json	off	Machine-readable result

serve

Flag	Env / default	Description
--host	MECHA_HOST (127.0.0.1)	Bind host
--port	MECHA_PORT (8080)	Bind port
--reload	off	Uvicorn auto-reload

describe

mechaharness describe lmstudio

Prints JSON from InferenceStrategy.describe() (name, base_url, model, auth).