HaShiShark

Cursor uses AI to edit code — we use AI to edit AI's context. 🪆 Context map + compression + version control for LLM context windows.

17
0
100% credibility
Found May 03, 2026 at 17 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

hashcode is a desktop app that visualizes an AI chat's context window as an editable map, using a second AI to compress, rewrite, or delete parts like a code editor.

How It Works

1
🔍 Discover hashcode

You hear about this cool desktop app that lets AI tidy up its own memory like editing code.

2
💻 Download and launch

Grab the app, install it with a quick click, and open your new AI workspace.

3
🔗 Link your AI helper

Pick your favorite AI chat service and connect it so everything works smoothly.

4
💬 Start chatting

Talk to your main AI while watching a live map of exactly what it remembers from the conversation.

5
🪆 AI edits its memory

Spot heavy parts on the map and let a second AI smartly trim, summarize, or delete them.

🎉 Smarter long chats

Your AI stays sharp and focused even after hours of back-and-forth, saving time and tokens.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 17 to 17 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 context-editor-agent?

This Python desktop app (Electron + React) visualizes your LLM chat's context window as a scrollable map with token-weighted nodes, letting a secondary AI agent inspect, compress, rewrite, or delete bloated items like tool outputs. It versions every edit for easy rollback, turning the black-box context into editable source code. Inspired by Cursor—where AI edits your code—this flips it to AI editing AI's context, supporting OpenAI, Claude, Gemini, and custom endpoints.

Why is it gaining traction?

Unlike blunt /compact commands or opaque tools like cursor use claude code, it offers a minimap, multi-select nodes, and precise tools (delete_item, compress_nodes) via a workbench with Suggest/Manual/Restore tabs. Developers dig the real-time streaming, file attachments, and mix-match providers—use Claude for context edits in cursor github integration setups. The hook: spot why your agent forgets mid-task and surgically fix it.

Who should use this?

AI agent builders wrestling long-context bloat in cursor github copilot flows or cursor github pull request reviews. Prompt engineers debugging cursor use outdated performance data—how to fix by pruning nodes. Teams mixing cursor use claude with cursor uses kimi, needing visibility into cursor uses which llm model without cursor github bad credentials errors.

Verdict

Alpha with 14 stars and 1.0% credibility score—rough edges, but solid docs and Windows installer make it playable now. Try for agent prototypes; contribute if context engineering hooks you.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.