agentbreeder

Quickstart

Zero to running AI agents in one command.

Two commands. Three minutes. Done.

pipx install agentbreeder
agentbreeder quickstart

No pipx? Use the standard installer instead

python3 -m pip install agentbreeder

python3 -m pip works on every Python install regardless of whether pip, pip3, or neither is on your PATH. If you see agentbreeder: command not found after, see Troubleshooting.

We recommend pipx because it installs agentbreeder into an isolated environment and puts the executable on PATH automatically — no shell-rc edits, no clashes with other tools.

Before you start

AgentBreeder needs three things on the machine:

  • Python 3.11+ (we recommend installing via pyenv)
  • One container runtime — Docker Desktop, OrbStack, Colima, Podman, or nerdctl
  • macOS 12+ · Linux · or Windows 11 + WSL2 · ~8 GiB free disk · ~4 GiB RAM

Not sure? Run agentbreeder doctor after installing — it checks every prerequisite, exits in under two seconds, and prints copy-pasteable fix commands for whatever's missing.

What to expect

agentbreeder quickstart is an interactive wizard. It asks a handful of questions (container runtime, model source, API keys, port conflicts) and pulls ~3 GB if you choose local models. Expect ~3 minutes on a cold first run.

Want zero prompts for CI? Use the scripted path:

agentbreeder quickstart --yes --no-ollama --no-browser

--yes skips every yes/no prompt and uses safe defaults. --no-ollama keeps you on the cloud path (set OPENAI_API_KEY etc. before running). --no-browser skips opening Studio.

agentbreeder quickstart is interactive and walks you through everything:

  1. Detects (or installs) Docker, Podman, or nerdctl
  2. Asks how you want to run models — Local, Cloud, or Both.
    • If Ollama is already running with models, Local is the default (press Enter — free, no key).
    • Otherwise the default is Both (install Ollama + add cloud keys).
  3. Sets up the path you chose:
    • Local — installs Ollama if needed, pulls gemma3 (~2 GB). No API key required.
    • Cloud — prompts for OpenRouter first (one key, 100+ models — recommended), then optionally for direct OpenAI / Anthropic / Google keys behind a single yes/no gate.
    • Both — local Ollama plus cloud keys.
  4. (macOS only, if Ollama is bound to 127.0.0.1) Optionally rebinds Ollama to 0.0.0.0:11434 so the Docker stack can reach it via host.docker.internal. This is opt-in (default: skip) — it restarts Ollama. Skipping is fine if you're using cloud or gateway models.
  5. Frees any conflicting ports for you on confirmation
  6. Starts the full stack — API · Studio · Postgres · Redis · ChromaDB · Neo4j · MCP servers · LiteLLM
  7. Seeds RAG docs, the knowledge graph, MCP servers, and prompts
  8. Deploys 5 sample agents (assistant, rag-agent, graph-agent, search-agent, a2a-orchestrator)
  9. Opens Studio at http://localhost:3001

Already have a cloud API key?

Skip the ~3 GB Ollama download with the --no-ollama flag:

agentbreeder quickstart --no-ollama

Equivalent to choosing Cloud at the model-source prompt. You'll be asked for OpenRouter first (one key, 100+ models), then optionally for OpenAI / Anthropic / Google direct keys.

Ollama and the Docker stack

Docker containers reach your host machine via host.docker.internal — not localhost. If Ollama is bound to 127.0.0.1 only (the macOS default), the AgentBreeder API container can't reach it even though ollama run works fine in your terminal.

Quickstart detects this and offers to rebind Ollama to 0.0.0.0:11434 (opt-in). You can also do it manually:

launchctl setenv OLLAMA_HOST 0.0.0.0:11434
osascript -e 'tell application "Ollama" to quit' && open -a Ollama

If you skip the rebind, agents that need local models will fall back to cloud providers.

Default login

Email: admin@agentbreeder.local · Password: plant

Studio will force you to set a new password on first sign-in — the default is publicly documented and unsafe to leave in place.


After it finishes

agentbreeder chat assistant            # talk to an agent
agentbreeder list agents               # see what's deployed
agentbreeder down                      # stop everything
agentbreeder quickstart --reset        # nuke and start fresh

Deploy to cloud from the same setup:

agentbreeder quickstart --cloud aws    # also: gcp | azure

Build your own agent

agentbreeder init                      # scaffold agent.yaml + code
agentbreeder validate agent.yaml
agentbreeder deploy agent.yaml --target local   # also: aws | gcp | azure | kubernetes

The 8-step deploy pipeline (parse → RBAC → resolve deps → build container → provision → health check → register → return endpoint) is atomic. If anything fails, the whole deploy rolls back.


Useful flags

FlagWhat it does
--yes / -yNon-interactive mode for CI. Skip every yes/no prompt and use safe defaults. Pair with --no-ollama and --no-browser for a fully scripted bootstrap.
--no-ollamaSkip the Ollama install/start/pull bootstrap
--ollama-model NAMEPull a different default model (e.g. llama3.2, phi4-mini)
--no-browserDon't open Studio at the end
--skip-seedSkip seeding sample data (faster restart)
--resetTear down all volumes and start fresh
--devBuild API/Studio images from source instead of Docker Hub
--skip-doctorSkip the prerequisite preflight (advanced; CI environments)

When something goes wrong

ProblemFix
agentbreeder: command not foundpip's script directory isn't on PATH. Either re-install with pipx install agentbreeder (recommended), or find the dir with python3 -c "import sysconfig; print(sysconfig.get_path('scripts'))" and prepend it to PATH in your shell rc. Running agentbreeder welcome once it works prints the same hint.
Port conflict (:5432 already in use etc.)Quickstart will offer to kill the conflicting process. Accept with y, or stop it manually.
Container daemon not runningQuickstart will tell you what to start (Docker Desktop, OrbStack, Colima, Podman, systemd, etc.) and re-detect after you confirm.
Stack already up, ports stuckagentbreeder down --clean then retry.
No LLM providerQuickstart prompts to install Ollama; or run agentbreeder setup separately for the full provider wizard.
Studio at :3001 is blankStale rajits/agentbreeder-dashboard:latest cached locally. docker rmi it, then agentbreeder quickstart --dev to rebuild from source. Full steps: FAQ →
migrate-1 shows ExitedExpected — migrate is a one-shot job that runs alembic upgrade head and exits. Only worry if the exit code is non-zero.
DOCKER_HOST points at a dead socketQuickstart detects this and points at the missing path. Run unset DOCKER_HOST (and remove the export line from your shell rc).
permission denied on /var/run/docker.sock (Linux)Your user isn't in the docker group. Quickstart prompts you to run sudo usermod -aG docker $USER && newgrp docker.
Snap / Flatpak Docker — "cannot connect to daemon"Quickstart detects the socket-path mismatch and tells you the exact export DOCKER_HOST=unix://<path> to run.
Docker socket reachable but docker CLI missingPartial install. Quickstart prints the distro-specific install command (brew install docker, apt-get install docker-ce-cli, etc.).
Rootless Docker volume permission errorsQuickstart prints a one-liner heads-up when rootless is detected. Bind-mounts are owned by your user, not root.

Next

You want to…Go here
See every CLI commandCLI reference →
See every agent.yaml fieldagent.yaml reference →
Deploy to AWS / GCP / Azure / KubernetesHow-To: Deploy →
Add knowledge bases (RAG / GraphRAG)RAG → · GraphRAG →
Add tools / MCP serversMCP servers →
Visual builderNo Code →
Python / TypeScript SDKFull Code →

On this page