glommer

glommer / cachebro

Public

File cache with diff tracking for AI coding agents. Drop-in MCP server that cuts token usage by 26%.

150
13
100% credibility
Found Feb 18, 2026 at 120 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

cachebro provides file caching with change detection for AI coding agents, returning full content on first read and compact diffs or unchanged notices on subsequent reads to save tokens.

How It Works

1
🔍 Discover cachebro

You learn about cachebro, a smart helper that makes AI coding assistants read files without wasting time or money on repeats.

2
📥 Set it up easily

You run one simple command to add cachebro to your coding editor, like Claude or Cursor.

3
✨ Editor gets smarter

Restart your editor, and now it has a memory that remembers files your AI has seen before.

4
🤖 Start coding with AI

Ask your AI helper to work on your project, and it automatically uses the smart file reading.

5
⚡ Saves time and cost

On repeat reads, it gets quick updates like 'no change' or just the differences, skipping full files.

6
📊 Check your savings

Look at the stats to see how many tokens were saved during your session.

🎉 Faster coding wins

Your AI works quicker and cheaper, letting you focus on building great projects.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 120 to 150 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 cachebro?

cachebro is a TypeScript-based file cache server designed for AI coding agents, slashing token usage by 26% on repeated file reads. It acts as a drop-in MCP server—agents call tools like read_file or read_files, getting full content first, then "unchanged" confirmations or compact diffs on changes, even for partial ranges. Powered by an embedded database, it runs locally with no network, handling file cache on Linux, Windows, or anywhere Node.js does, and integrates via simple npx cachebro init for editors like Cursor or Claude Code.

Why is it gaining traction?

Agents discover and prefer its MCP tools automatically from descriptions promising token savings, no special prompts needed—unlike manual file github copilot instructions or basic read tools. Benchmarks on real TypeScript refactors show 26% fewer tokens with identical results, compounding over tasks, while CLI status tracks savings and github file history-like diffs avoid full re-reads. Easy setup beats file cache conflict in IntelliJ or generic file cache services, with batch reads and force-refresh for control.

Who should use this?

AI agent users in Cursor, Claude Code, or OpenCode building on codebases with github file size limits or frequent edits. Devs running Claude Opus on Node/Python projects, Houdini pipelines, or GitHub Actions needing file cache service for agents without token burn. Teams tracking file github download or history in multi-session workflows.

Verdict

Try it if you're deep in agent-driven coding—solid docs, CLI, and SDK make it practical despite 113 stars and 1.0% credibility score signaling early maturity. Low risk via npx, but watch for edge cases like large github file size limit 100mb files or custom timeouts.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.