Engram

Engram keeps project knowledge between coding sessions so developers don't have to repeat past decisions, bug fixes, and conventions to their AI agent. It works with existing coding assistants through MCP.

• Searchable project memory: Agents can save discoveries and retrieve relevant context through full-text search, session summaries, and timelines, with reads scoped to the current project by default.
• Agent compatibility: Use the same memory system with Claude Code, Codex, Gemini CLI, OpenCode, VS Code Copilot, Cursor, Windsurf, or other MCP-compatible agents.
• Memory review: Browse and search saved memories in a terminal interface, and let agents inspect related or conflicting entries before deciding what applies.
• Sharing across machines: Git sync transfers memories as portable compressed files. Optional Engram Cloud replicates selected projects to a self-hosted server and gives teams a browser dashboard for shared memory and activity.

Engram is open source under the MIT license and runs on Windows, macOS, and Linux. The local tool is a single Go binary with no separate runtime dependencies; it stores memories in a SQLite database with built-in FTS5 search. Developers can also access that memory through an HTTP API or CLI.

Memory stays local by default, and the local database remains authoritative when replication is enabled. Engram Cloud runs on your own infrastructure using Docker and PostgreSQL; only projects you explicitly select are replicated to that server.