- Who it helps
- Developers investigating persistent memory for coding assistants.
- Availability
- Public Apache-2.0 repository checked November 1. The documented memory engine needs Python and no model key; separate research dependencies are larger.
What is available
The engine provides command-line access and a Model Context Protocol server, which lets compatible assistants call its tools. Records go into memory/events.jsonl. A supersede operation can mark an earlier fact as replaced while retaining its history.
What remains experimental
The author distinguishes the working log and retrieval tools from the unproved adaptive-memory design. The README records negative experiments and a repaired evaluation corpus. It says the planned comparison against compaction and vector retrieval still lacks required evaluation components.
What you can do
- Read the current research status and storage format before trying a disposable project. Open reference ↗
Where the evidence stops
Documentation checked; no installation or retrieval test executed. Local storage does not determine what a connected cloud assistant receives.
Sources read. Primary sources read November 1, 2026 UTC. No third-party installation, inference or complete workflow executed.
Sources & further reading
GitHub · official
Offline project memory keeps revisions inspectable ↗
Identifies the subject, documented capability or specific verification limit.
krapcys1-maker / GitHub · page checked 2026-11-01
Coverage note: Independent attention and end-to-end execution remain unverified.
The record over time (1)
Dated source checks and published explanations. An observation date can differ from the original release date.
- · Added to Current
An experimental local memory engine stores project decisions in an append-only log and exposes search and revision tools.