Skip to content
AGENT-LAB

Search the documentation

Type to search every page of the documentation.

Open source · MIT · Python 3.12+

Agent-Lab

Build, simulate and test LLM agents.

A cloud-native Python framework on FastAPI, LangGraph and PostgreSQL.

Get startedRead the docs
$ uvx cookiecutter gh:btech-software/agent-lab --directory cookiecutter
$ cd my-agent-app && docker compose up --build -d
$ curl http://localhost:18000/status/liveness
# REST API at /agents · MCP server at /mcp · OpenAPI docs at /docs

Build

Subclass a base agent, register it with one decorator, call create_app(). REST, MCP, persistence and auth come wired in.

Developer's GuideMCP Server

Simulate

Run agents against live LLMs with a simulated user and an LLM-as-judge, tracked over time in Langfuse and LangWatch.

Researcher's GuideSimulations

Test

Validate every change against real infrastructure, spun up by testcontainers. No mocks where it matters.

Testing guide

Tutorial · about 15 minutes

From zero to a running, observable agent.

  1. Scaffold a project from the template.
  2. Run the full stack with Docker Compose.
  3. Talk to your agent over REST, then from an MCP client.
Your first agentDeveloper's Guide
# 1 · scaffold
$ uvx cookiecutter gh:btech-software/agent-lab --directory cookiecutter
# 2 · install and test with no infrastructure
$ cd my-agent-app
$ uv lock && uv sync --group dev --group test
$ make test_unit
# 3 · run the full local stack
$ docker compose up --build -d
$ curl http://localhost:18000/status/liveness

Pick your path

The documentation is organized by what you are trying to do.