Basic Memory

Basic Memory is an open source knowledge base and memory layer for AI tools, built around the Model Context Protocol. Conversations, decisions, notes and research are stored as plain Markdown files that form a knowledge graph, and a full text, semantic and graph index sits on top so that an AI assistant can search and traverse that knowledge rather than reloading whole files. The pitch is persistent memory that the user owns and can inspect, instead of the small, vendor controlled memory built into chat products. It suits individuals and teams who work across several AI tools and want the same context everywhere. It connects to Claude, ChatGPT, Gemini, Codex, Cursor, VS Code and Obsidian through MCP. Notes link through wiki links and typed relations, can be filtered by frontmatter fields such as status, type, tags and priority, and can be validated against schemas for note types like Task, Person, Client or Ticket. Because everything is Markdown, files can be opened in any editor, versioned with git and backed up however the user prefers. The local app is free and open source under the AGPL version 3.0 license and runs offline on the user's own machine. Cloud plans add private sync across desktop, mobile and web, and each starts with a seven day trial. Team costs $15 per seat per month, falling to $13 per seat at 10 or more seats and $12 at 20 or more, and includes shared workspaces, roles and version history. Business costs $30 per seat per month with SSO and SAML, audit logs, admin controls, HIPAA compliant hosting, priority support and invoiced annual billing. Agent Infrastructure is custom, with credit based usage for agent fleets and on premises options. The product is built by Basic Machines, founded by Paul Hernandez, and an MSP partner program offers consolidated billing for managing multiple clients.