AgentLantern

AgentLantern is a suite of developer tools for multi-agent LLM systems.

Point it at any agent project — it detects the framework automatically, documents the architecture, checks design issues, and can visualize supported crews while they run.

---

Lantern Tools

AgentLantern is organized around a small CLI surface. Each command does one job and can be used from the root of an agent project or with an explicit project path.

01

Document the system

Generate a structured docs site with overview, architecture, diagrams, agents, tasks, runbook, and contacts.

lantern web

02

Check the design

Run deterministic lint rules for missing fields, broken references, idle agents, YAML issues, and CI output.

lantern lint

03

Watch the run

Start, stop, inspect, and replay supported agent runs inside the animated Play environment.

lantern play

lantern web      # generate and serve docs for the current project
lantern lint     # inspect project quality without LLM calls
lantern play     # open the live animated run viewer

---

Supported Frameworks

Framework	lantern docs	lantern lint	lantern play
CrewAI	✅ Full analysis	✅ 13 rules	✅ Live runtime
Google ADK	✅ Full analysis	Planned	Planned
LangGraph	Detected	Planned	Planned
AutoGen	Detected	Planned	Planned
Smolagents	Detected	Planned	Planned

Framework detection is automatic — no configuration needed.

---

Installation

pip install "agentlantern @ git+https://github.com/brellsanwouo/agentlantern.git"
# or (recommended)
uv tool install git+https://github.com/brellsanwouo/agentlantern.git

Verify:

lantern --help

See the full Installation guide.