jayminwest

jayminwest / mulch

Public

Growing Expertise for Coding Agents — structured expertise files that accumulate over time, live in git, work with any agent

ai-memory ai-tools claude-code
99
15
100% credibility
Found Feb 17, 2026 at 22 stars 4x -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

Mulch stores structured project knowledge like conventions, failures, and decisions in a git-tracked folder, allowing AI agents to load relevant expertise before tasks and record new insights after.

How It Works

1
🏠 Discover Mulch

You hear about Mulch, a simple way to help AI helpers remember important project lessons so they get smarter each time.

2
🌱 Set up in your project

With one easy step, you create a special spot in your project to store growing wisdom.

3
📁 Add knowledge areas

You name categories like 'database tips' or 'api tricks' to organize what you learn.

4
💡 Save your first insight

You quickly note a useful rule, mistake to avoid, or smart choice, and it sticks forever.

5
🤖 Boost your AI helper

Your AI gets a perfect summary of all project know-how to work faster and better.

6
👥 Share with teammates

Everyone on the team pulls the latest smarts automatically when they update their project.

🌟 Knowledge compounds

Over time, your project builds endless wisdom, making AI helpers and humans super effective together.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 22 to 99 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is mulch?

Mulch is a TypeScript CLI for coding agents to build persistent, git-tracked expertise in repos. Agents record structured insights—like conventions, failures, patterns, or decisions—into domain-specific JSONL files via `mulch record`, query them with `mulch query`, and prime session context using `mulch prime`. This solves agents forgetting patterns between sessions, enabling growing expertise that compounds across runs, teammates, and providers without any LLM dependency.

Why is it gaining traction?

In agent-driven dev, where tools like Claude or Cursor reset context each time, mulch stands out as a passive, git-native layer for what enables growing expertise—append-only records with BM25 search, compaction, and multi-agent locking for safe concurrent writes. Its provider hooks (Claude, Cursor, Aider) and commands like `mulch learn` for changed-file suggestions make growing my expertise seamless, turning one-off discoveries into shared, searchable knowledge.

Who should use this?

AI coding teams at startups or open-source projects using agents for repetitive tasks like debugging APIs or migrations. Ideal for backend devs growing expertise in e-health knowledge and skills, or any group tired of re-explaining project quirks to new sessions.

Verdict

Try mulch if agents are core to your workflow—early at 18 stars and 1.0% credibility score, but solid docs, tests, and MIT license make it low-risk for green growing.expertise centers. Maturity lags big repos, so pair with `mulch doctor` for health checks.

(187 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.