Skip to content

Trace Your LangChain Agent in 2 Minutes

Already using LangChain? Add tracing in 2 lines.

Install

The [langchain] extra only pulls langchain-core. To run a full LangChain agent you'll also need langchain itself plus a model adapter such as langchain-openai:

pip install "fastaiagent[langchain]" langchain langchain-openai

Add Tracing to Your Existing Code

# Your existing LangChain code - unchanged
from langchain_openai import ChatOpenAI
from langchain_core.messages import HumanMessage

llm = ChatOpenAI(model="gpt-4.1")

# Add these 2 lines
import fastaiagent
fastaiagent.integrations.langchain.enable()
handler = fastaiagent.integrations.langchain.get_callback_handler()

# Run as usual — pass the handler via callbacks
response = llm.invoke(
    [HumanMessage(content="What's the weather?")],
    config={"callbacks": [handler]},
)

LangChain 1.x

Earlier versions of this tutorial used langchain.agents.create_tool_calling_agent and AgentExecutor. Those symbols were removed in LangChain 1.x — use langchain.agents.create_agent (or langgraph) instead.

Traces are stored locally in SQLite. View them:

fastaiagent traces list --last 24h
fastaiagent replay <trace_id>

What the Handler Hooks

The callback handler subclasses langchain_core.callbacks.BaseCallbackHandler and instruments:

  • on_llm_start / on_llm_end — langchain.llm.<model> span
  • on_tool_start / on_tool_end — langchain.tool.<name> span
  • on_llm_error / on_tool_error — closes the matching open span on failure

Spans land in .fastaiagent/local.db and are visible through fastaiagent traces list and the replay/export CLI commands.

Disable Tracing

fastaiagent.integrations.langchain.disable()

Push Traces to the Platform

from fastaiagent import FastAI

fa = FastAI(api_key="sk-...", project="my-project")
# Traces are automatically pushed when platform is connected

Next Steps