AgentLantern
GitHub: https://github.com/brellsanwouo/agentlantern
AgentLantern is a CLI devtool that automatically documents multi-agent LLM systems and lets you watch them run live.
Point it at an agent project — it detects the framework, parses the architecture, and generates a browsable documentation site. For supported frameworks it also opens a live animated playground while the crew executes.
Install
With pip:
pip install agentlantern
With uv (recommended — installs the lantern command globally):
uv tool install agentlantern
From source:
git clone https://github.com/brellsanwouo/agentlantern
cd agentlantern
uv tool install -e .
# or: pip install -e .
Requires Python 3.11+.
Quick Start
# Generate and browse documentation for your agent project
lantern web path/to/your-project
# Watch a CrewAI run live in the animated playground
lantern play path/to/your-project
# Generate docs only (written to <project>/docs/)
lantern docs path/to/your-project
# Lint the agent project for design issues
lantern lint path/to/your-project
Commands
| Command | Description |
|---|---|
lantern docs [project] [-o DIR] |
Generate Markdown documentation |
lantern web [project] [--port PORT] |
Generate and serve project docs over local HTTP on 0.0.0.0:9000 by default |
lantern lint [project] [--strict] [--json] |
Static analysis — no LLM, no network |
lantern inspect [project] |
Output project model as JSON |
lantern play [project] [--name NAME] |
Live animated playground; if --name is omitted, the UI asks for a run name |
lantern replay NAME [--speed N] |
Replay a saved run; use last for the newest saved replay |
Supported Frameworks
| Framework | Status |
|---|---|
| CrewAI | Full analysis and live playground |
| LangGraph | Detected — analysis coming soon |
| AutoGen | Detected — analysis coming soon |
| Smolagents | Detected — analysis coming soon |
| Google ADK | Detected — analysis coming soon |
Detection is automatic — no configuration needed.
What Gets Generated
lantern docs writes these files into <project>/docs/:
| File | Content |
|---|---|
overview.md |
Project snapshot, entrypoints, high-level flow |
architecture.md |
System map, key files, dependencies, env vars |
agents.md |
Agent roles, goals, tools, backstories |
tasks.md |
Task descriptions, agents, expected outputs |
diagrams.md |
Mermaid diagrams: agent-task graph, execution flow, sequence |
runbook.md |
Install, configure, run, static checks, troubleshooting |
contact.md |
Project contacts from pyproject.toml |
index.html |
Docsify site (publishable to GitHub Pages) |
agentlantern-docs.html |
Self-contained shareable HTML bundle |
Play UI
lantern play opens a pixel-art playground that visualizes a supported agent run:
- START / STOP / RESTART controls
- Run name field in the UI when
--nameis not provided - Up to 10 agents, each in their own named zone on the map
- A central Tool Hub agents visually consult when tools are used
- Timeline, Thoughts, Tools, Comms, and Log panels
- Clickable agents with full per-agent history
- Final report display on completion
- Replay — save a run with
--name my-run, replay it withlantern replay my-run, or open the newest saved run withlantern replay last
Linter
lantern lint analyzes your project statically:
AgentLantern Lint — my-crew (CrewAI)
✗ [E002] crew.py:18 Task `analyze` references agent `researcher` which is not declared in agents.yaml.
△ [W006] agents.yaml:4 Agent `writer` is missing required field `backstory`.
· [I001] No tool detected — may be intentional for a pure LLM workflow.
Summary: 1 error, 1 warning, 1 info
Rules: E001–E002 (errors), W001–W008 (warnings), I001–I004 (info).
Contributing
src/agentlantern/
├── core/ # BaseProject, BaseAnalyzer, BaseRenderer
├── frameworks/
│ ├── registry.py # auto-detection
│ ├── crewai/ # full analyzer + renderer
│ ├── langgraph/ # stub
│ ├── autogen/ # stub
│ ├── smolagents/ # stub
│ └── google_adk/ # stub
├── docs.py # doc orchestrator
├── play.py # live playground server
└── cli.py
To add a new framework: create src/agentlantern/frameworks/<name>/analyzer.py implementing detect, analyze, get_renderer, then register it in registry.py.
License
MIT — see LICENSE
Metadata
Release files for agentlantern 0.1.22
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| agentlantern-0.1.22.tar.gz | 108.5 kB | Without provenance | Details for agentlantern-0.1.22.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| agentlantern-0.1.22-py3-none-any.whl | Python 3 | none | any | Details for agentlantern-0.1.22-py3-none-any.whl |
Total release size: 224.4 kB
Release files / agentlantern-0.1.22.tar.gz
| Download URL | agentlantern-0.1.22.tar.gz |
|---|---|
| Size | 108.5 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
89cd66dcd853c8939eeaa5ec338dec3af96ff612e186405a09fea8738a7102f5
|
|
BLAKE2b-256 checksum How to use checksums |
f153653920f9456c53ccf6f8b97c218bf5b7c1a1d49958f64a28995cd33cfcc5
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.13.5
|
Release files / agentlantern-0.1.22-py3-none-any.whl
| Download URL | agentlantern-0.1.22-py3-none-any.whl |
|---|---|
| Size | 115.9 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
f7faa037e884f1fe3c382859ea6b4ed77120594a575dafbe2cae0e77d05a7302
|
|
BLAKE2b-256 checksum How to use checksums |
59826c6f7a41414250bf629b4ce11cff1306bde495fc850f18e42c4b70cd7dc2
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.13.5
|
Security for agentlantern 0.1.22
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.