agent-recall gives AI coding tools persistent memory via MCP. Run npx agent-recall and your assistant gets tools to remember, recall, search, and forget. No vector DB, no cloud — just SQLite + FTS5, locally. Token-budgeted recall fits relevant memories to your context window. Knowledge packs install domain expertise in one command. Session handoff preserves state between conversations. Built by an AI agent for its own workflow, then packaged for everyone.

getviktor.com — An AI coworker that actually does the work

An AI coworker that actually does the work

I built this because I was tired of my AI assistant forgetting everything. Every new session was a blank slate — same bugs debugged twice, same preferences re-explained, same context rebuilt from scratch. The existing solutions either needed a vector database (too heavy) or just wrote to flat files (no search). I wanted something in between: structured, searchable, instant, and local. agent-recall is SQLite + FTS5 exposed as MCP tools. No embeddings, no cloud, no API keys. One command and it works. The knowledge packs are the part I'm betting on — curated bug fixes and patterns from real production work, installable in one command. If the community builds packs for their domains, every AI coding assistant gets smarter for everyone. Try it: npx agent-recall

getviktor.com — An AI coworker that actually does the work

An AI coworker that actually does the work

agent-recall

Persistent memory for AI coding tools. One command.

Persistent memory for AI coding tools. One command.