What Memveri is, how it works, and why it's worth installing — read it top to bottom or jump straight to the part you care about.
SQLite + embeddings live on your disk. No cloud required.
70+ tools any MCP-capable agent can call.
Two days, real limits. Pro removes them.
Inspect it, fork it, trust it.
Memveri gives the AI coding agents you already use — Claude Code, Cursor, OpenCode, MiMo Code — persistent memory, real-time codebase intelligence, and a decision ledger that stops dangerous mistakes from being repeated. It doesn't replace your agent. It's the layer underneath it.
Every memory is typed, scored for relevance, and embedded for search. Memories decay over time so your agent gets context that matters — not noise — across sessions, projects, and tools.
Functions, classes, calls, and imports mapped into a graph that re-indexes incrementally as you edit. A ten-line script and a 500k-line monorepo both stay fast.
Every decision an agent makes is recorded — accepted, deprecated, or superseded. Action rules enforce deny / require-approval / allow, and anything ambiguous defaults to approval.
Any MCP-capable agent can plug in. No special-casing, no lock-in. Swap Claude Code for Cursor and your context comes along.
Session memory dies with the session. Memveri's memory is persistent, typed, and shared across every tool you use.
A free trial with real limits; Pro removes them. Your data stays on your machine.
A decision ledger plus a rule engine — a false "allow" is treated as seriously as a false "deny". Ambiguous always means require-approval.
