HutchDB
Store, query, and share persistent structured records across multiple AI agents without schemas or SQL.
HutchDB is an open-source, schema-optional database built specifically for AI agents and coding assistants like Claude Code, Cursor, and Codex. Operating as a headless MCP server powered by PostgreSQL JSONB, it replaces fragile markdown files with durable, individually addressable records that auto-create collections on first write. Multiple agents can collaborate on the same shared workspace without overwriting each other, utilizing rich MongoDB-style filtering, full-text search, and field aggregations across sessions.
主要功能
01Schema-optional JSONB storage with automatic collection creation on first write
02Headless, single-user deployment via Docker with zero setup ceremonies and optional bearer-token authentication
03Rich query capabilities featuring MongoDB-style operators, containment filters, and full-text search
04Multi-agent data sharing allowing different MCP clients to read and update identical records
05Built-in aggregation, schema inference, custom views, and CSV/JSON import/export tools
0624 GitHub stars
使用案例
01Accumulating long-term research like competitor analysis or customer interviews across multiple agent chats
02Orchestrating agent handoffs in multi-stage data processing or lead enrichment pipelines
03Tracking project backlogs, bug reports, and key architectural decisions that persist past context window limits