MCP/HutchDB
HutchDB logo

HutchDB

Headless agent-native database for AI agents. Store, query, and share persistent structured JSONB records across sessions without schemas or SQL.

rag-data•Databases & Storage
GitHub22
License:AGPL-3.0
VERIFIED
ByExpedited Projects@Expedited Projects

About this Server

HutchDB is a headless, single-user database and Model Context Protocol (MCP) server built specifically for AI coding agents. It gives multiple agents (Claude Code, Cursor, Codex, Claw) shared, durable persistence across sessions using schema-optional Postgres JSONB with full-text search, Mongo-style operators, aggregations, and saved views—with zero OAuth ceremony or dashboards.

Real-time Interaction

Direct low-latency RPC protocol communication with structured JSON schemas.

Scoped Sandboxing

Restricted capability bounds to prevent accidental modifications or leakage.

Security & Best Practices

API Key Enforcement: When hosting off localhost, configure HUTCH_API_KEY to require bearer authentication on every call.
Network Isolation: Keep the Docker container inside an internal private bridge network or behind a reverse proxy.
Data Sanitization: Ensure sensitive system secrets or private keys are not stored in shared agent collections.

When to Use (and When Not To)

Best For
  • •Cross-session persistence and shared agent memory across multiple AI assistants and editors
  • •Storing unstructured user interview transcripts, scratch notes, task queues, and research findings
  • •Schema-less record prototyping without writing SQL migration files or running schema migrations
Not For / Limitations
  • •High-dimensional vector embedding similarity search (use Chroma MCP instead)
  • •Strict multi-table relational foreign key constraints and ACID joins across engines (use DBHub MCP instead)
  • •Production enterprise Elasticsearch cluster monitoring (use Elasticsearch MCP instead)

Install & Setup

terminal command
$claude mcp add hutch --transport http http://localhost:3000/api/mcp

Exposed Tools & Endpoints

6 tools

Functions your AI agent can invoke via this Model Context Protocol server

hutch_store

Stores records in a collection without pre-defining schemas or table structures.

hutch_query

Queries records with Mongo-style comparison operators, projections, sorting, and aggregations.

hutch_search

Performs full-text keyword search across stored JSONB record contents in a collection.

hutch_update

Updates existing records matching specific filter criteria or IDs.

hutch_delete

Deletes specific records or collections matching query conditions.

hutch_list_collections

Lists all active collections, record counts, and storage statistics.