Agent-Lab is a cloud-native Python framework for building, simulating and testing LLM
agents on top of FastAPI, LangChain/LangGraph and PostgreSQL. pip install btech-agent-lab, extend the agent base classes, and assemble a production-ready
application with a single create_app() call — REST API, MCP server, persistence,
auth and OpenTelemetry observability included.
The documentation is organized by what you are trying to do, in the spirit of the Diátaxis framework: tutorials when you are learning, guides when you are working, reference when you need a fact. Pick your path:
| I want to… | Path |
|---|---|
| Get started | Your first agent → Developer's Guide |
| Build an application | Developer's Guide → Configuration reference → REST API reference → MCP Server guide |
| Evaluate and research | Researcher's Guide → Testing & Simulations |
| Deploy and operate | Deployment guide → Observability → Keycloak → Vault |
| Contribute | Developer's Guide → Entity Domain Model → Acknowledgements |
Tutorials
Learning-oriented: walk through end-to-end, don't deviate.
| Page | Description |
|---|---|
| Your first agent | Scaffold a project, define an agent, call it over REST, watch the trace, expose an MCP tool — in ~15 minutes. |
Guides
Working-oriented: solve the problem you have right now.
| Page | Description |
|---|---|
| Developer's Guide | The flagship guide: two development models, configuration, DI, building agents, every extension point. |
| MCP Server | Expose agents to MCP clients (Claude Desktop, MCP Inspector); add tools and prompts. |
| Testing & Simulations | Integration suite with testcontainers; LLM-as-judge agent simulations with Langfuse/LangWatch. |
| Researcher's Guide | Jupyter notebooks, experiments, and prototyping new agent behaviors. |
| Deployment | Local Docker Compose stacks and production Kubernetes with Terraform + Helm. |
| Auth with Keycloak | Realm, client and service-account setup; token flows the API expects. |
| Secrets with Vault | How configuration secrets are stored, retrieved and rotated. |
| Observability with Grafana · with OpenSearch | Reference dashboards for the two bundled observability stacks. |
Reference
Facts only: exact, complete, generated from the code.
| Page | Description |
|---|---|
| Configuration reference | Every config key and environment variable, and how the active ConfigSource is chosen. |
| REST API | The full endpoint catalog with request/response examples. |
| Entity Domain Model | Entities, relationships and the three PostgreSQL databases. |
Project
| Page | Description |
|---|---|
| Acknowledgements | Contributors and third-party software. |
| License | MIT. |
What's next?
New to Agent-Lab? Start with Your first agent — a complete, copy-paste walkthrough that ends with a running agent you can talk to over REST and MCP.