Jev in production › Agent context and memory

jevmory

Local memory tool for Claude Code and MCP agents in which Jev grades each MEMORY.md line against stored session evidence for support and contradiction, labelling it stale, wrong, unsupported or keep, with each judgment stored as a receipt.

Open on GitHub ↗

The source publishes no measured number.
Use
Agent context and memory
Industry
Developer tools
Form
Open-source tool
Stage
In production
Plugs into
Claude Code
Listed
2026-09-22
Found via
github
Repository
romiluz13/jevmory
Stars
11
Forks
1
Last push
2026-10-06
Language
Python
License
MIT

The README opens with

Local, zero-dependency memory layer for coding agents. Every remembered fact is a verbatim quote graded by TypeSafe Jev's calibrated confidence — and every number has a receipt.

Your coding agent forgets everything at session end. The usual fix — a hand-maintained MEMORY.md — rots silently: stale lines, lines nobody ever said, lines the project outgrew, and no way to tell which is which. jevmory closes both ends: it remembers what was actually said (hooks ingest every session locally), and it audits the memory file against that evidence, verdict by verdict.

Badge

For the project's own README, linking back here:

Listed in Jev in production

Also used for agent context and memory