Shared Memory for AI Agents

A lightweight database that lets AI agents from different providers share context and data

Problem

Developers now routinely use multiple agents—Claude, Codex, Cursor—and each one keeps its own context. Information gathered by one agent can't be reused by another, so project background and decisions have to be re-fed each time. Someone built a dedicated database for AI agents to store and share context, confirming that cross-agent context silos are a real pain point.

Solution

A local-first lightweight database/service with a standard API and MCP interface that lets AI agents from different providers read and write the same structured context (project background, decision logs, code snippets). It supports natural-language queries and vector search, with data stored locally or self-hosted so agents pull relevant memory on demand.

Users

Developers using multiple AI coding agents, small teams needing long-term project memory, self-hosting AI workflow enthusiasts

MVP

Local SQLite/vector storage; MCP server integration with major agents; basic read/write API with namespace isolation; natural-language query; one-click export/import; runs with minimal setup

Revenue

Open-source core + hosted sync and team sharing ($8/user/mo); enterprise self-hosted license; integration partnerships with model providers

Why now

Multi-agent workflows are becoming the default for developers, but cross-agent context sharing has no established standard; the maturing MCP protocol lets indie developers plug in via an adapter layer instead of defining their own protocol