Quickstart
:::info Reference version This is the guided path: a local Ollama model, step by step, with checkpoints. The product quickstart is the concise reference for the same stack. :::
What you will build
You will run the complete OpenJCockpit stack on your own machine and sign in to the AI Orchestration Dashboard. At the end you see the landing hub with the Business and Implementation portals.
- Docker with Docker Compose
- Ollama on the host (not in Docker), with enough memory for a large model
- An
OPENAI_API_KEY. It is mandatory to boot the stack, because thebestandcheapestaliases and embeddings are still OpenAI-hosted. - Free host ports
4000,4403-4406,4100,8080,8091,9080and8181 - Git, to clone the
OpenJCockpitandOpenJCockpit-frontendrepositories side by side
develop (draft, not yet verified on a clean machine). OpenJCockpit is in active development, so details may differ on other versions.qwen3.6:27b). Output varies from run to run and is slower than hosted models, so your results will not match the screenshots exactly. Long waits in the video are sped up and labelled.Steps
1. Pull the model and keep Ollama running
ollama pull qwen3.6:27b
ollama list
ollama list shows qwen3.6:27b. Keep the Ollama
daemon running while the stack is up.
2. Clone both repositories next to each other
git clone https://github.com/OpenJCockpit/OpenJCockpit.git
git clone https://github.com/OpenJCockpit/OpenJCockpit-frontend.git
3. Configure and start the backend
cd OpenJCockpit
cp .env.example .env
# edit .env and set OPENAI_API_KEY
docker compose up --build
All other variables in .env.example are commented out and have working local defaults. Use a
throwaway key for tutorials and never commit .env.
The build finishes and the one-shot local-seed container exits
successfully. It prepares Keycloak roles, local users and the LiteLLM keys,
and it is safe to run again.
:::caution Health is not readiness for inference
/actuator/health reports UP even when Ollama is unreachable. A missing model only shows up when
the first LLM call fails. See Troubleshooting.
:::
4. Start the dashboard
In a second terminal:
cd OpenJCockpit-frontend
docker compose up --build
The dashboard proxies /api to ai-control-service on port 9080 and signs in against Keycloak on
8080.
5. Sign in
Open http://localhost:4000. You are redirected to the OpenJCockpit login page. Sign in with a
seeded local user.
The seeded users and their passwords are listed in the product
authentication guide.
Use tony, who also has the openjcockpit-admin role.
After login you land on the hub: an intro greeting with your first name, a top bar, and the Business and Implementation portals.
6. Pull the job image once
Workflow runs on a project use a container image that is not pulled automatically:
docker pull eclipse-temurin:25-jre
What just happened
Docker Compose started Keycloak (identity), PostgreSQL, the LiteLLM gateway, OPA (policy) and the Java services. The dashboard is a separate application that talks to them over host ports. Ollama serves the coding model for every agent. For the architecture, see the product overview.
Stop, restart and reset
The runbooks describe every service, volume and destructive command.
| Command | Effect |
|---|---|
docker compose stop | Stops containers, keeps all state |
docker compose down | Removes containers, keeps named volumes |
docker compose down -v | Destructive. Deletes the realm, databases and keys |
Troubleshooting
See Troubleshooting for the most common snags.
Next steps
Your first spec-driven change runs a real workflow on the demo project.