tracehawk 0.1.0

TraceHawk SDK — observability for AI agents

TraceHawk Python SDK

Observability for AI agents. Automatically instrument OpenAI, Anthropic, MCP, and LangChain calls and send traces to TraceHawk.

Install

pip install tracehawk

Quick start

import tracehawk
from openai import OpenAI

tracehawk.init(
    api_key="th-...",          # from TraceHawk Settings → API Keys
    app_name="my-agent",
)

client = OpenAI()

@tracehawk.observe
def run_agent(prompt: str) -> str:
    response = client.chat.completions.create(
        model="gpt-4o",
        messages=[{"role": "user", "content": prompt}],
    )
    return response.choices[0].message.content

All OpenAI, Anthropic, and LangChain calls inside @tracehawk.observe are automatically traced — no manual span creation needed.

Configuration

Parameter	Env var	Default
api_key	TRACEHAWK_API_KEY	—
endpoint	TRACEHAWK_ENDPOINT	https://app.tracehawk.dev/api/otel
app_name	—	tracehawk-app
capture_content	—	True
disable_batch	—	False

MCP tracing

pip install tracehawk[mcp]

MCP tool calls are automatically captured as MCP-kind spans with mcp.tool_name attributes.

Self-hosted

tracehawk.init(
    api_key="th-...",
    endpoint="http://localhost:3001/api/otel",
)

License

Apache 2.0