AgentLens Documentation

Observability and Explainability for AI Agents — see inside the black box.

As AI agents become more autonomous — making decisions, calling tools, chaining actions — you need to see what they're doing and understand why. AgentLens is an open-source observability platform that gives you full visibility into agent behavior.

Why AgentLens?

How It Works

1. Instrument — Add the Python SDK to your agent. Use decorators or manual tracking.
2. Collect — Events are batched and sent to the Express.js backend over HTTP.
3. Store — Events and sessions are persisted in SQLite (zero external dependencies).
4. Visualize — The built-in dashboard shows timelines, token charts, and explanations.

Quick Example

Pythonimport agentlens
from agentlens import track_agent, track_tool_call

# Initialize
agentlens.init(api_key="my-key", endpoint="http://localhost:3000")
session = agentlens.start_session(agent_name="my-agent")

# Decorate your functions — tracking is automatic
@track_tool_call(tool_name="web_search")
def search(query: str) -> list:
    return do_search(query)

@track_agent(model="gpt-4")
def my_agent(prompt: str) -> str:
    results = search(prompt)
    return synthesize(results)

# Run and explain
my_agent("What's the weather in Seattle?")
print(agentlens.explain())
agentlens.end_session()

Components