Give agents a database that speaks MCP out of the box
FlowDB
AI-native database stack with MCP out of the box
The situation
The problem
Agents need a place to put memory, documents, and structured records — then retrieve them without a custom integration for every framework. Most databases were built for apps, not for tools an agent can call.
How it usually works
Builders glue a document store, a vector DB, and an MCP wrapper together, then hope the schema survives the next agent runtime. Data for agents often ends up in a hosted RAG vendor by default.
How we ran it
FlowDB is an open-source, AI-native database stack: JSON in, JSON out, automatic vectorization for RAG, MCP compatibility out of the box, and a REST API. You run it where you want — including next to Courier.
Strategy
Make the data layer agent-native instead of retrofitting a human database. MCP should be a ship feature, not a weekend project.
Implementation
NoSQL-native storage, vectors on ingest, MCP and a client SDK, Dockerized so one command runs it anywhere. Builders keep the data on infrastructure they choose.
Capabilities
JSON In-and-Out
NoSQL-native storage. Send JSON, get JSON. No schema migrations, no ORM headaches.
Auto Vectorization for RAG
Documents are automatically vectorized on ingest, making retrieval-augmented generation a first-class citizen.
MCP Out-of-the-Box
Ships with MCP support and agent compatibility. Connect FlowDB to any AI assistant or agent framework without custom integration work.
API Controlled & Client SDK
Full REST API and client SDK for programmatic control. Manage databases, collections, and queries entirely through code.
Dockerized Deployments
One command to run. FlowDB ships as a Docker image so you can deploy anywhere with zero configuration.