butterbase-ai/butterbase

Butterbase is an open-source backend that handles the infrastructure most apps need, databases, user authentication, file storage, serverless functions, and AI capabilities, so you can focus on building your product instead of wiring up backend plumbing. It's designed for AI-driven applications and includes a built-in AI gateway that routes requests to language models, plus managed retrieval-augmented generation (RAG) for semantic search over your documents. The distinguishing feature is that every backend capability is exposed as tools that AI agents can use directly, so an agent can operate your backend without glue code. At a high level, the project bundles several services into one stack: a Postgres database split across three isolated planes for platform metadata, runtime state, and per-app user data, file storage backed by S3-compatible services, a key-value store, and a runtime for executing TypeScript functions. There's also a Python-based agent runtime for long-running tasks. You interact with all of it through a single API that auto-generates REST endpoints from your database schema, handles auth flows, and exposes everything through a Model Context Protocol (MCP) server, meaning AI assistants like Claude can call your backend's tools directly. The project also ships a Claude Code plugin with 30+ guided skills for scaffolding and deploying apps. This would appeal to founders or product teams building AI-powered apps who want a backend without the overhead of assembling separate services. For example, someone building a chat app could use the durable actors for stateful chat rooms, the realtime subscriptions for live updates, and the AI gateway for LLM calls, all from one platform. A team building a document-search tool could use the managed RAG collections for ingestion and semantic search. The repo is deliberately split between open source and a managed offering. The open-source version includes the full runtime, everything needed to self-host, but leaves out multi-region orchestration, billing logic, upstream AI provider integrations, and admin dashboards. Self-hosters get unlimited quotas with no-op billing and can implement their own providers through documented interfaces. The tradeoff is that the open-source distribution is still young, and you'll need some comfort with Docker and command-line setup to run it locally.