mattpocock

mattpocock / sandcastle

Public

Orchestrate sandboxed coding agents in TypeScript with sandcastle.run()

31
1
100% credibility
Found Mar 26, 2026 at 31 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

Sandcastle lets everyday people run AI coding helpers on their projects in safe isolated spaces, automatically blending the AI's improvements back into the main work.

How It Works

1
🔍 Discover Sandcastle

You find this helpful tool while looking for ways to let an AI safely improve your project.

2
📦 Add it to your project

With one simple command, you bring Sandcastle into your work folder.

3
⚙️ Set it up quickly

Run a setup command that creates a special folder with everything needed.

4
🔑 Connect your AI helper

Add your AI service details to a simple settings file so the magic can begin.

5
✏️ Describe the task

Write plain instructions in a note about what changes you want the AI to make.

6
🚀 Launch the AI agent

Hit run and watch the AI work safely in its own space, step by step.

✅ Enjoy the results

Your project gets updated with clean changes, ready to use right away.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 31 to 31 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 sandcastle?

Sandcastle is a TypeScript library for running AI coding agents in isolated Docker containers, handling git worktrees, agent invocation, and commit merging automatically. You get AFK agents that edit your repo locally via a simple `run()` call or CLI, with flexible prompts pulling dynamic context like open GitHub issues. Unlike remote tools like IBM Watsonx Orchestrate GitHub or Cesium Sandcastle GitHub, it's 100% local and git-native.

Why is it gaining traction?

It skips sync headaches with bind-mounted worktrees, letting agents write directly to your host filesystem—no manual push/pull. Templates kickstart workflows like sequential reviewers or parallel planners, and prompts support `!` commands for repo state plus `{{}}` args for reuse. Devs dig the programmatic API for CI/scripts over rigid platforms like orchestrate ADK GitHub.

Who should use this?

Repo owners automating GitHub issue fixes with Claude agents, TS/JS teams prototyping sandcastles AI for code gen, or indie devs needing local sandcastle builder GitHub alternatives to escape vendor lock-in. Ideal for sandcastles in the sand-style experiments without cloud costs.

Verdict

Try it for agent-driven coding if you're in TypeScript/Docker—solid docs, templates, and tests make setup fast despite 31 stars and 1.0% credibility score. Early maturity means watch for edge cases, but it's a practical step past project sandcastle GitHub toys.

(187 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.