What is komi-learn?
komi-learn is an open-source memory layer for AI coding agents that learns your work patterns and improves its own skills across sessions. It plugs into Claude Code and Codex and provides continuous, zero-friction memory: relevant learnings load automatically at the start of a session based on current context, without you running manual commands. After a session, background distillation extracts durable lessons from the transcript, and a curation step merges overlapping lessons and archives outdated ones over time so the knowledge base stays clean. A privacy-filtering step removes secrets, machine-specific paths, and one-off failures before anything is processed. An optional community pool lets users share general lessons through a GitHub-based repository with content-addressed and signed contributions. In practice, komi-learn learns your coding styles, technology stacks, and effective fixes so the agent performs better each time you use it. It is aimed at developers using AI coding agents who want consistent improvement across sessions. The project is MIT-licensed and free, distributed on PyPI and requiring Python 3.10+, with model access via a Claude Code login or API key.
What is komi-learn used for?
Developers who use AI pair-programming tools daily use komi-learn to give Claude Code and Codex persistent memory of their stack and coding style across sessions.
How much does komi-learn cost?
Free (Open source)
Open-source package installable from PyPI, running locally at no cost.
Pros and cons
Pros
• Gives AI coding agents persistent memory across sessions • Runs locally in the background, observing project structure, stack, and coding patterns • Injects context automatically at session start — no slash commands or manual exports • Works out of the box with Claude Code and OpenAI Codex • Open source, so it is auditable and extensible • Runs entirely on your machine, so nothing leaves your environment
Cons
• Distributed via PyPI, so it assumes a Python environment and CLI comfort • Supports Claude Code and OpenAI Codex specifically, not every AI coding tool • Local background process adds another component to run and maintain • Value is mainly for developers who use AI pair-programming tools daily
Questions about komi-learn
What problem does komi-learn solve?
It fixes AI coding agents forgetting everything when a session ends by giving them persistent memory of your project across sessions.
How does komi-learn work?
It runs locally in the background, observing your project structure, stack choices, and coding patterns, then injects that context automatically when your next session starts.
Which AI coding tools does it support?
komi-learn is compatible with Claude Code and OpenAI Codex out of the box.
Does komi-learn send my code to the cloud?
No. It runs entirely on your machine, so nothing leaves your environment. It is also open source and auditable.
Do I need to run commands to use it?
No. Context is injected automatically at session start — no slash commands, manual exports, or cloud sync required.
What are the best alternatives to komi-learn?
Reviews
Sign in to rate and review komi-learn.
Sign in to reviewNo reviews yet. Be the first to review komi-learn!