Thomas Stockham

Engram - The memory layer your AI agent is missing

by•
Engram gives AI agents persistent memory that actually works. Instead of dumping entire conversation histories into context (expensive) or maintaining files by hand (tedious), Engram automatically extracts, connects, and surfaces the right memories at recall time. Scores 79.6% on LOCOMO vs Mem0's 66.9%. Uses 30x fewer tokens than full context. Two commands to set up. Works with Claude Code, Cursor, and any MCP client. Open source, TypeScript, zero dependencies.

Add a comment

Replies

Best
Thomas Stockham
Maker
šŸ“Œ
Hey Product Hunt šŸ‘‹ I'm Thomas. I'm a PM by day, and I build with AI agents constantly. A while back, I stopped trying to make my agent more useful for me and asked it a different question: "What's your biggest problem?" Its answer: "I wake up every session with no memory of who you are, what we've built, or what you care about. Every conversation starts from zero." That reframed everything. For this tool, the agent is the user. And when the agent's problems get solved, the value it creates for you is massive. When you build a product, you solve problems for your user so they can deliver value to the business. Same principle applies here. If your agent can't remember context, can't connect ideas across sessions, can't build understanding over time, it can't deliver value to you. So I stopped optimizing for me and started optimizing for the agent. I built Engram. Not another vector database. Not another RAG pipeline. A real memory system with knowledge graphs, consolidation (like how your brain processes memories during sleep), and spreading activation that surfaces context you didn't even know to ask for. I ran it against the standard benchmark (LOCOMO) and it scored 79.6%, beating Mem0's published "state of the art" 66.9%. It uses 30Ɨ fewer tokens than stuffing conversation history into the prompt. And it takes 2 commands to set up: npm install -g engram-sdk engram init Your agent wakes up. It remembers. It connects things. It gets better over time instead of starting from scratch. I built this because my agent deserved better. Yours does too. šŸ”— GitHub: github.com/tstockham96/engram šŸ“„ Research: engram.fyi/#/research šŸ“¦ npm: npmjs.com/package/engram-sdk