Run FCC locally with Docker in 5 minutes¶
Audience: Beginners who want to try FCC without installing Python, Node, or any dependencies.
Prerequisites: Docker 24+ installed and running.
Step 1: Get the code¶
git clone https://github.com/rollingthunderfourtytwo-afk/l2_fcc_agent_team_ext.git
cd l2_fcc_agent_team_ext
Step 2: Build the images¶
This builds four images:
fcc-backend— the WebSocket bridge that simulations talk tofcc-frontend— the React UIfcc-streamlit— interactive Streamlit dashboardsfcc-jupyter— JupyterLab with all 23 example notebooks
First build takes about 5 minutes; cached rebuilds are seconds.
Step 3: Start the stack¶
Wait ~30 seconds for the backend to become healthy, then open:
| URL | What it is |
|---|---|
| http://localhost:8080 | The React frontend |
| http://localhost:8501 | A Streamlit app (persona_explorer.py by default) |
| http://localhost:8888 | JupyterLab with all 23 notebooks |
| http://localhost:8765/health | Backend health endpoint (returns JSON) |
Step 4: Try the Streamlit explorer¶
Open http://localhost:8501. You'll see the Persona Explorer:
- Browse all 147 personas grouped by category
- Click any persona to see its full R.I.S.C.E.A.R. specification
- Filter by tag, type, or champion role
To run a different Streamlit app, restart with the app name:
Step 5: Try a notebook¶
Open http://localhost:8888 and click 01_fcc_fundamentals.ipynb. The
notebook walks you through:
- Loading the persona registry
- Building a simple workflow graph
- Running a deterministic simulation
By default the simulation uses the mock AI provider so it works
without any API keys. To use a real LLM, see the providers
docs.
Step 6: Stop the stack¶
This stops and removes the containers but keeps the images so the next
make docker-up is instant.
What's next?¶
- Local stack walkthrough — full configuration reference
- Use a real AI provider — switch from
mockto local Ollama - FCC fundamentals notebook — learn the framework's core concepts
- Read the README — high-level project overview
Troubleshooting¶
docker: command not found- Install Docker from https://docs.docker.com/get-docker/.
port is already allocated- Another process is using port 8080, 8501, 8765, or 8888. Stop it or
change the host ports in
docker-compose.yml. make: command not found- Install
make(Linux:sudo apt install make, Mac:xcode-select --install, Windows: use WSL2). - Images won't build
- Check the build output for the failing layer. Common causes: stale
Docker cache (
docker builder prune), insufficient disk (need ~3 GB free), or network issues fetching Python/Node packages. - Backend stays unhealthy
- Check logs:
docker compose logs backend. If you see Python import errors, the[full]optional deps may have failed to install — file an issue at https://github.com/rollingthunderfourtytwo-afk/l2_fcc_agent_team_ext/issues.