HTTP API reference

App: mechaharness.api.app:app
Stability: evolving (v1)
Python helper: mechaharness.factory.run uses the same RunRequest / RunResponse (mechaharness.core.contract).

Routes

Method	Path	Description
GET	/health	Liveness + version
GET	/backends	Configured inference backends
GET	/families	Configured harness families
POST	/v1/run	Run a prompt through a harness

POST /v1/run

Request body

Field	Type	Notes
prompt	string	Required
backend	string?	Defaults from settings
family	string?	Defaults from settings
model	string?	Defaults from settings
api_key	string?	Optional override
base_url	string?	Optional override
system_prompt	string?	Optional
max_turns	int	Default 8
temperature	float?	Optional

Response body

Field	Type	Notes
final_text	string?	Model’s final answer
turns	int	Loop iterations used
messages	object[]	Full transcript
events	object[]	Turn / tool observability events

Example

curl -s http://127.0.0.1:8080/v1/run \
  -H 'content-type: application/json' \
  -d '{
    "prompt": "What is 2+2?",
    "backend": "lmstudio",
    "family": "tool_loop",
    "model": "local-model"
  }'