Installation

Install reagent-flow and choose the adapter path that fits your stack.

Core package

Install the zero-dependency core first:

pip

pip install reagent-flow

uv add reagent-flow

Framework adapters

Install only the adapters you need:

pip

pip install reagent-flow-openai
pip install reagent-flow-anthropic
pip install reagent-flow-langchain
pip install reagent-flow-langgraph
pip install reagent-flow-crewai

uv add reagent-flow-openai
uv add reagent-flow-anthropic
uv add reagent-flow-langchain
uv add reagent-flow-langgraph
uv add reagent-flow-crewai

Which path should you use?

Core Only

Start with manual session, log_llm_call, and log_tool_result instrumentation.

Adapter Path

Auto-capture traces from OpenAI, Anthropic, LangChain, LangGraph, or CrewAI.

See the Demo

Jump straight to the flagship multi-agent showcase.

Verify the install

Create a quick smoke test:

import reagent_flow

with reagent_flow.session("smoke-test") as s:
    s.log_llm_call(tool_calls=[{"name": "ping", "arguments": {}}])
    s.log_tool_result("ping", result={"ok": True})

s.assert_called("ping")

Run it with pytest:

uv run pytest test_smoke.py -v

If that passes, the core install is working. For CI setup, see CI.