About Boat House
Boat House is open-source infrastructure for teams building small apps with AI agents. It provides hosted or self-hosted deployment, team access, persistent data, and versioned updates through a CLI, MCP endpoint, and agent skill.
Key features
- 01Deploy apps with HTTPS addresses
- 02Team sign-in and per-tool access roles
- 03Per-tool databases and persistent files
- 04Versioned publishing, rollback, and backups
- 05CLI and agent skill
Use cases
- 01Publish an agent-built app for team use
- 02Share an app with invited editors or viewers
- 03Update or roll back a deployed app
MCP support
Boat House includes an MCP endpoint that works alongside its `bh` CLI and agent skill for deploying and managing apps.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://mcp.boathousecloud.com/mcpclaude mcp add --transport http com-boathousecloud-boathouse https://mcp.boathousecloud.com/mcp{"mcpServers":{"com-boathousecloud-boathouse":{"type":"http","url":"https://mcp.boathousecloud.com/mcp"}}}Open in Cursor{"servers":{"com-boathousecloud-boathouse":{"type":"http","url":"https://mcp.boathousecloud.com/mcp"}}}Open in VS Code{"mcpServers":{"com-boathousecloud-boathouse":{"type":"http","url":"https://mcp.boathousecloud.com/mcp"}}}{"mcpServers":{"com-boathousecloud-boathouse":{"serverUrl":"https://mcp.boathousecloud.com/mcp"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
https://mcp.boathousecloud.com/mcpAuth requiredTools
41| Parameter | Type | Required | Description |
|---|---|---|---|
| workspace | string | No | Workspace slug to act in. Use this when you belong to several workspaces; whoami lists them. |
Details
From the project READMEBoat House
Open-source infrastructure for small software built with AI agents.
Boat House gives a tool a live HTTPS address, sign-in, a database, persistent files, and sharing by email. Editors use the app in a browser. Developers with Admin access can use their own agents to fetch its source and publish versioned updates to the same address.
The software is licensed under Apache 2.0. You can run it on your own infrastructure or use Boat House managed hosting.
| Self-host | Managed hosting | |
|---|---|---|
| Software | Free under Apache 2.0 | Same open-source foundation |
| Infrastructure and operations | You provide and maintain them | Boat House operates them |
| Price | Your server, domain and provider costs | $10 per organization per month for up to five lightweight tools; domain purchases are separate |
| Getting started | Self-hosting guide | Sign up, confirm email, copy the connection into your agent |
What is included
- A FastAPI control plane and Caddy HTTPS gateway.
- Account sign-in, email invitations, per-tool Viewer / Editor / Admin access, and signed identity passed to apps.
- Per-tool PostgreSQL databases, persistent file volumes, container resource limits, and saved source releases.
- Deployment, source checkout, updates, rollback, export, and backup/restore commands.
- The
bhCLI, an MCP endpoint, and an agent skill for Claude Code, Codex and other compatible agents. - Optional email, domain registration, and prepaid billing integrations. Self-hosters configure their own providers; metering can be disabled.
Admin is full control of a tool, including its source, secrets, sharing and deletion. Software collaboration uses source checkout and versioned publication. App code enforces its own record-level access rules.
Start with managed hosting
Create an account at boathousecloud.com, confirm your email, then choose Copy connection and paste the message into your coding agent. Ask it to put a tool online or share it by email. Add credit before deploying. No separate hosting, database or sign-in project is needed.
Install the agent skill in Cursor, Codex or Windsurf:
npx skills add allenzhangyiteng/boathouse-skills
For Claude Code, add /plugin marketplace add allenzhangyiteng/boathouse-skills, then /plugin install boathouse@boathouse. See the skill and plugin repository and installation guide.
The CLI can create a working app before you connect an account:
bh init my-app --template team
cd my-app
python3 app.py --local
Use --template static for an HTML website. boathouse is an alias installed alongside bh when available. The starter format is ordinary source plus boathouse.json; no proprietary runtime is needed. See the runnable guides, llms.txt and complete agent docs.
Run your own installation
See the self-hosting guide for a Linux host with Docker, DNS and HTTPS. The public release contains no accounts, credentials, database snapshots, hosted customer tools or production configuration. All examples use placeholders; generate your own secrets.
This is infrastructure for trusted teams and trusted app builders. The control plane manages the host Docker daemon. Resource limits and container networks do not make it a secure sandbox for hostile code. Read the security model before exposing an installation.
Development
Python 3.12 is used by the container image. In a virtual environment:
python -m pip install -r control/requirements.txt -r requirements-dev.txt
python -m pytest -q
python scripts/check_public_tree.py
Tests create temporary local state and use fake providers. Do not point tests at production credentials or state. CI runs the tests, publication hygiene checks, and Gitleaks.
Repository layout
| Path | Purpose |
|---|---|
control/app/ | API, account UI, gateway authorization, deployment and providers |
control/site/ | Public-site source and static assets |
cli/bh | Standard-library Python CLI |
skill/boathouse/ | Agent instructions |
caddy/, docker-compose.yml | Gateway and services |
deploy/ | Host installation, synchronization and backup scripts |
examples/hello/ | Minimal app that checks signed identity |
tests/ | Regression tests with synthetic data |
Contributions are welcome; see CONTRIBUTING.md. Report vulnerabilities using SECURITY.md.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.