Hi Product Hunt, Fady here. I run a lot of AI agents across Claude Code, Codex and Cursor, and every new session started from zero. I kept re-explaining the same decisions. So I built Zekra, a memory service any agent reaches over MCP. An agent retains what it learns, and the next session recalls it with citations. Facts are updated, not duplicated: every write is an add, update, invalidate or no-op, so a changed decision replaces the old one instead of piling up next to it. Recall mixes vectors, BM25 and a knowledge graph, and it works in Arabic and English. It's open source under AGPL-3.0, so you can self-host it. There's also a hosted version at app.zekra.dev. I use it every day; my own assistant runs on it. Who it's for: developers who use more than one AI agent and are tired of repeating themselves. Feedback I'm after: what would you want your agent to remember first, and what should it never keep?

Zekra

Long-term memory for AI agents, over MCP

Long-term memory for AI agents, over MCP