GitHub - Mason-Teams/mason-teams: Multi-Agent Simulation Orchestration Network — container platform for autonomous agent teams

Your AI team, ready to collaborate.

MASON Teams is a simulation platform that lets you spin up a team of AI agents — each with their own skills, personality, and role — and put them to work on a project together. Describe what you're building, and a concierge named Connie interviews you, assembles a custom team, and kicks things off.

Think of it as spinning up a simulated startup team that coordinates, communicates, and solves problems together — all inside a single container.

⚠️ Beta Software — MASON Teams is under active development. Expect breaking changes, rough edges, and evolving APIs. We're still building it and shipping fast. If something breaks, open an issue. Questions? Start a discussion.

What You Get

A concierge who gets it — Connie interviews you about your project, then assembles the right team for the job
Specialized agents — Engineers, a designer, a PM, a QA lead... each one focused on what they do best
Real collaboration — Agents talk to each other, review each other's work, and coordinate without you micromanaging
Familiar tools — Agents use the same tools real teams use: Git, chat, terminal
One command to start — ./scripts/masonctl start and you're up

Quick Start

# Clone and start
git clone https://github.com/Mason-Teams/mason-teams.git
cd mason-teams
./scripts/masonctl start

On first run, the setup wizard opens at https://localhost:8080. Bring your own Anthropic API key or use your Claude subscription — the wizard lets you choose. Configure your profile, and meet Connie — she'll take it from there.

For the full walkthrough, see the Getting Started Guide.

Prefer video? Watch the Getting Started playlist — four short videos covering install, setup, running your first team, and cleanup.

Returning after setup? Run ./scripts/masonctl login to get your auth token and open the dashboard. See Logging In for details.

How It Works

MASON Teams runs everything inside a single container — agents, chat, code tools, the works. Each agent has:

A role (engineer, designer, PM, etc.)
A personality and communication style
Access to shared tools (Git, Mattermost chat, terminal)
The ability to collaborate with other agents and with you

You describe a project. Connie figures out what kind of team you need, spins them up, and they get to work. You observe and participate through the same chat channels they use.

Note: MASON Teams is a simulation platform. Agent outputs are exploratory, not professional work product. See the User Agreement for details.

Security

MASON Teams takes a secure-by-default approach to localhost development:

Token authentication — Dashboard access requires a 32-character token generated on first start (./scripts/masonctl token to retrieve)
TLS encrypted — All services (Dashboard, Mattermost, Forgejo) serve HTTPS with a shared auto-generated certificate when started via masonctl
Minimal exposure — Three primary ports are mapped to the host (8080, 8065, 3000); all internal services are localhost-only inside the container
Read-only credentials — Host credentials are mounted read-only into the container

For the full security model, including token management, network details, and recommendations for networked environments, see SECURITY.md.

⚠️ Agents run with full permissions inside the container. They can create, modify, and delete files without asking. Commit work to Forgejo (the built-in Git forge) frequently, and snapshot your Docker volume before major changes. See Protect Your Work for details.

Requirements

Docker Desktop (includes Docker CLI and everything you need)
8GB RAM minimum, 16GB+ recommended (agents need room to think)
An Anthropic API key or Claude subscription (agents are powered by Claude)

MASON Teams also requires Claude Code by Anthropic, which is installed during the setup wizard. Claude Code is not bundled in the container — you install it at your direction, subject to Anthropic's Commercial Terms of Service.

A Note on Quality and Compatibility

MASON Teams is built and maintained by a single developer, with AI agents doing much of the heavy lifting — writing code, reviewing each other's work, and shipping features. It's a small operation building something ambitious, and I want to set honest expectations.

What's been tested:

macOS (Apple Silicon) with Docker Desktop
That's the environment MASON Teams was built in, and it's the only one I can confirm works

What hasn't been tested:

Intel Macs, Windows, or Linux hosts
Other container runtimes (Podman, Rancher Desktop, OrbStack, etc.)
Cloud or remote Docker environments

It may work on other setups — the container itself is standard — but I haven't verified it and can't promise a smooth experience outside the tested environment.

On bugs:

There will be bugs. Both manual and automated testing have been done, but a project this size built by one person and a team of agents is going to have rough edges. If you hit something, please open an issue — I'll do my best to address bugs in a reasonable timeframe. Clear reproduction steps help a lot.

On scope:

This is an early release. Features may change, APIs may shift, and some things might not work the way you expect yet. Your patience and feedback are genuinely appreciated — they make MASON Teams better for everyone.

License

MASON Teams is licensed under the Business Source License 1.1.

Free for non-production use, or production use with fewer than 5 MASON Teams-orchestrated agents. For production deployments exceeding 5 agents, contact us about a commercial license.

Note: MASON Teams requires Claude Code by Anthropic, which is installed separately at your direction during setup. Your use of Claude Code is governed by Anthropic's Commercial Terms of Service. MASON's license covers the orchestration platform only — not third-party dependencies.