---
title: For builders
type: landing
for: [developers, agent-builders]
---

The AI memory you wanted when you tried to vibe-code your own.

You hit the context limit. You built a folder of notes and some glue. This is the finished version.

Plain Markdown you own, with a real index on top. Full-text, semantic, and a knowledge graph. Open source, MCP-native, runs local.

## You've tried this

We know, because we did too.

You dumped chats into a folder. Maybe you wired up a vector DB, or threw together an MCP server, or kept a sprawling CLAUDE.md. It half-worked. Until it got big, or you switched tools, or you couldn't find anything. Basic Memory started exactly there. A side project born from beer and bitching about this exact problem. What is Basic Memory?

## What you actually wanted

Files you can still read. That the agent can actually use.

Not a black box you can't inspect, and not a folder the agent has to grep. You wanted plain files you own, with real search and a graph on top, that show up in every tool you work in. That's the part that's annoying to build. So we built it.

## That's Basic Memory

The finished version of the thing you started.

Plain text + a real index

Markdown is the source of truth; full-text, semantic, and graph search make it queryable. Read it like a doc, query it like a database.

MCP-native, every tool

One memory across Claude Code, Cursor, Codex, and any MCP client. Your agent reads context before it writes code.

Open source, runs local

AGPL-3.0, SQLite, local embeddings, no API keys required. Self-host it, or sync a private cloud when you want to.

## Free, open, and yours

Run it local. Read the code. Fork it if you want.

Basic Memory is AGPL-3.0. The whole stack lives on GitHub: the MCP server, the CLI, the indexer, the schemas, the integration plugins, the bundled skills. pip install basic-memory or brew install basic-memory and you have the same engine the cloud runs on, on your laptop, with no API keys and no phone-home.

## Drop it into your stack

Two minutes to connect your first tool.

Then follow a playbook: persistent memory for coding or durable memory for your agents.

Stop rebuilding it.

Free and open source to run locally. Two minutes to connect your first AI tool.