oxly 0.1.0

Chrome DevTools for AI Agents — observability SDK for LangGraph, CrewAI, AutoGen

Oxly SDK

Python instrumentation for Oxly — real-time tracing, security analysis, cost tracking, and Time Machine replay for AI agents built on LangGraph, CrewAI, AutoGen, or custom Python.

Oxly runs as one FastAPI process backed by SQLite — no Redis, no ClickHouse, no separate Collector or Workers to deploy. This SDK talks directly to that process over HTTPS, with a local SQLite fallback if it's unreachable.

Install

pip install oxly

Quickstart

1. Run the Oxly API (see the main README for the full walkthrough — no Docker, Redis, or ClickHouse required):

git clone https://github.com/ramakrishna1967/Oxly
cd Oxly
pip install -e packages/api
export JWT_SECRET_KEY=$(openssl rand -hex 32)
export DATABASE_URL="sqlite+aiosqlite:///oxly.db"
uvicorn api.main:app --host 0.0.0.0 --port 8000

2. Instrument your agent:

from oxly import observe, init

init(
    collector_url="http://localhost:8000",  # your running Oxly API
    api_key="ak_your_project_key",          # from POST /api/v1/projects
)

@observe
def research_agent(query: str) -> str:
    context = search_tool(query)
    return llm.chat(f"Answer based on: {context}")

@observe(name="planning.step")
async def async_agent(objective: str) -> list[str]:
    return await llm.achat(f"Break this into steps: {objective}")

Every call produces a full trace — arguments, return values, timing, exceptions, token counts, and cost — visible in the dashboard at http://localhost:8000.

Zero-interference guarantee. @observe will never crash your application. If Oxly hits an internal error, your function runs normally and the SDK fails silently.

Configuration

init() accepts explicit kwargs, or falls back to OXLY_* environment variables (OXLY_API_KEY, OXLY_COLLECTOR_URL, OXLY_ENABLED, OXLY_SERVICE_NAME, OXLY_DEBUG, ...). See oxly.config.OxlyConfig for the full list and defaults.

License

Apache 2.0