Your AI finally remembers

PLUR gives AI agents memory that persists. You correct Claude Code on Monday — Cursor remembers on Tuesday. Engrams live as plain YAML on your disk; recall is fully local; nothing leaves your machine unless you ask.

This site documents three things: the basic local install, the adapters that connect PLUR to your tools, and PLUR Enterprise for teams.

Get started What PLUR is, how it works, install in 30 seconds.

Concepts Engrams, episodes, scopes, activation, packs, sync.

Adapters MCP, OpenClaw, Hermes, CLI, custom integrations.

Reference Schema, MCP tools, CLI commands, REST API, benchmarks.

Guides Multi-project, sync, packs, feedback, migrations.

Enterprise Self-hosted, SSO, SCIM, audit, admin dashboard.

Proven

In our local-knowledge benchmark, agents with PLUR memory win 89% of decided contests across Haiku, Sonnet, and Opus. On the house-rules subset: 12 wins, 0 losses, every model. On LongMemEval (retrieval): 86.7% overall, 93.3% Hit@10 — with zero API calls and zero per-query cost.

The headline that surprises people: Haiku with PLUR memory outperforms Opus without it. The bottleneck isn’t model intelligence — it’s context. Benchmark methodology →

How it works

You correct your agent   →  engram created    →  YAML on your disk
Agent fixes an incident  →  episode captured  →  timeline searchable
Next session starts      →  relevant engrams injected →  agent remembers
You rate the result      →  engram strengthens or decays →  quality improves

Two storage primitives — engrams (typed assertions that decay if unused) and episodes (timestamped events). Search is fully local: BM25 + BGE embeddings + Reciprocal Rank Fusion. No cloud, no API costs, works offline.

License

Apache-2.0. PLUR is open source, built by the Fair Data Society, used by teams that want their agents to remember without sending their knowledge anywhere.