A fleet manager
for your AI agents
Deploy agents in isolated Docker containers, interact with them via a web dashboard, and extend them with plugins. Fully self-hosted.
Fleet Dashboard
Everything you need to run AI agents
Built for engineers who want full control over their AI infrastructure.
Docker isolation
Every agent runs in its own container. Strict resource limits, network isolation, and clean restarts by default.
Plugin system
Extend agents with MCP tools via independent Docker containers. Install by pointing at a manifest URL.
Agent Packs
Portable zip files that define an agent's persona, tools, and skills. Share, version, and deploy in seconds.
Up and running in minutes
Three steps from zero to a live agent fleet.
Configure
Copy the example env file and set your encryption key, access token secret, and refresh token secret.
cp apps/api/.env.example apps/api/.env Deploy
Run Docker Compose. On first boot, NanoFleet prints a temporary password and a QR code in the terminal — scan it to set up 2FA.
docker compose up --build Deploy your first agent
Open the dashboard, create an agent, and write its SOUL.md and TOOLS.md directly from the UI. Enable plugins as needed.
# http://localhost:8080
Portable agents,
ready to deploy
An Agent Pack is a zip file you import from the dashboard. NanoFleet mounts it into the agent's container — persona, tools, and skills included.
- manifest.json — name, model, required API keys
- SOUL.md — persona, behavior rules, and constraints
- TOOLS.md — how the agent uses its MCP tools
- /skills — optional custom files
# Import from the dashboard → Deploy Agent Extend with plugins
Plugins are independent Docker containers exposing an MCP server. Install by pointing NanoFleet at a manifest URL.
Ready to deploy your fleet?
NanoFleet is open source and self-hosted. Your agents, your infrastructure.