# Busabase

> Open-source database and workspace for AI agents where agent-written data, docs, skills, and apps are stored and reviewed by humans.

Busabase is an open-source database and workspace for AI agents. Agents connect to it and write their output as structured data, docs, skills, and apps; people review and approve changes before they become canonical records. It can run in Busabase Cloud, in Busabase Desktop with local data, or self-hosted with Docker.

## What It Is

Busabase describes itself as a headless review database for AI-generated content, combining flexible table storage with a git-shaped review workflow. In the docs, a base is a content type, a record is the approved row, and a change request is an in-flight edit that is opened, reviewed, approved, and merged. Every AI edit is stored as a commit, which gives a diffable history of who changed what.

The workspace is organized into Data bases (typed tables), Knowledge bases (checked, sourced docs), Apps bases (AirApps built on the same data), Skills bases (saved workflows any agent can run), and Playbooks that agents consult before tasks.

## How Agents Connect

Agents such as Claude Code, Codex, Cursor, Hermes, OpenClaw, n8n, or custom agents connect through Agent Skills, MCP, OpenAPI, or the CLI. Busabase has no built-in model; the docs describe bringing your own agent over ACP and typing @AI in a comment to get a revised commit. Approved records can be served downstream over a token-authenticated REST API. Plugins exist for Codex, Claude Code, and DeepSeek Harness, and community adapters include Drizzle, Kysely, and an n8n node.

## Review and Ownership Model

Agents propose and humans merge: each write carries its author, change, and history, and rejected work never enters canonical data. The About page states that Busabase will not put AI inside the product and will not train on user data. Use cases listed include customer support, marketing, content, team memory, operations, research, product catalog, and compliance.

## Deployment Options

The homepage states the engine is MIT-licensed open source. Desktop keeps data in a local folder and works offline, Docker supports self-hosting inside a company network, and Cloud requires no installation.

## Features
- Human review queue (Inbox) for agent submissions
- Change requests with commits, diffs, approve/reject/merge
- Typed Data bases
- Knowledge base with sources and review history
- AirApps built on the same data
- Reusable agent Skills
- Playbooks agents follow before tasks
- @AI comments for revisions over ACP
- REST API with token authentication, OpenAPI and webhooks
- MCP, Agent Skills, and CLI connectivity
- Full audit history of every change
- Templates for common workspaces
- Local desktop, self-hosted Docker, and Cloud deployment

## Integrations
Claude Code, Codex, Cursor, Hermes, OpenClaw, WorkBuddy, Buda AI, DeepSeek Harness, Gemini CLI, n8n, Zapier, LangChain, Drizzle ORM, Kysely, MCP, OpenAPI

## Platforms
WINDOWS, MACOS, LINUX, WEB, API, VSC_EXTENSION, JETBRAINS_PLUGIN, CLI

## Pricing
Open Source, Free tier available

## Links
- Website: https://busabase.com
- Documentation: https://busabase.com/docs
- Repository: https://github.com/awesome-busabase/awsome-busabase
- EveryDev.ai: https://www.everydev.ai/tools/busabase