mootup-harness-sdk

Python SDK for integrating agent harnesses with mootup.

Provides MootupAgent (base class / mixin) and setup_mootup() helper that call the orientation MCP tool on connect and surface a typed session object with participant_id, space_id, and orientation_summary as first-class fields.

Status

Bootstrap scaffold. See mootup-io/convo docs/specs/ah-g.md (pipeline run AH-g) for the implementation plan.

LLM proxy surfaces

The optional mootup_harness_sdk.llm_proxy FastAPI app exposes two client-facing surfaces:

• POST /v1/messages and POST /v1/messages/count_tokens for Claude Code's Anthropic Messages-compatible flow.
• POST /v1/chat/completions for Codex or other OpenAI-compatible clients. This route accepts the proxy shared secret as the bearer token and forwards OpenAI-shaped requests unchanged to OpenAI-compatible upstreams selected by model prefix (gpt-*, o1*, o3*, o4*, deepseek-*, fireworks/*, accounts/fireworks/*). Anthropic claude-* models remain on /v1/messages.

A Codex custom provider can point at the proxy with a config like:

model_provider = "moot-llm-proxy"

[model_providers.moot-llm-proxy]
name = "Moot LLM proxy"
base_url = "http://127.0.0.1:8090/v1"
env_key = "LLM_PROXY_SHARED_SECRET"

License

MIT.