Sangkwun

Sangkwun / sandy

Public

Deterministic MCP scenario replay for AI agents — create once, replay without LLM inference

automation claude-code-plugin deterministic mcp scenario-replay
11
0
100% credibility
Found Feb 05, 2026 at 10 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

Sandy enables AI agents to record, parameterize, and replay sequences of actions deterministically to accelerate workflows and reduce costs.

How It Works

1
🔍 Discover Sandy

While chatting with your AI helper about repeating the same chores, it suggests Sandy to make future tasks lightning fast.

2
📱 Add the Helper

With one simple command in your AI chat app, you bring Sandy into your toolkit, and it's ready to go.

3
💡 Spot a Repeatable Task

Your AI finishes a job like grabbing website info and saving it, then smartly saves the perfect sequence of moves for next time.

4
⚡ Replay Instantly

The next time you ask for something similar, Sandy replays the saved steps in a flash without your AI rethinking everything.

5
🔄 Tweak and Reuse

Adjust details like names or links in the saved plan, and run it again smoothly for endless variations.

🎉 Work Smarter

Your AI helper now zips through repeated jobs, saving you time and money while feeling super reliable.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 10 to 11 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 sandy?

Sandy records MCP tool call sequences from AI agents as parameterized JSON scenarios, then replays them deterministically without LLM inference—create once for scraping, DB queries, or browser automation, and reuse forever. It eliminates token costs and timing variability for repeatable agent workflows, delivering consistent results every time. Python-based, it installs as a Claude Code plugin via simple marketplace commands or runs standalone with `play scenario.json --var KEY=VALUE`.

Why is it gaining traction?

Zero-inference replays make agents blazing fast for deterministic simulation testing github, outpacing LLM-heavy alternatives that rack up costs on loops. CLI flags like `--dry-run`, `--start N`, and `--include-results on_failure` let devs debug and customize effortlessly. MCP transport auto-detection hooks into Claude, Cursor, and stdio servers seamlessly, appealing to agents create enthusiasts dodging LLM inference overhead.

Who should use this?

Claude Code users automating browser flows or data pipelines tired of re-reasoning the same steps. AI agent builders doing deterministic policy gradient github training or MCP-heavy simulations. Devs scripting repetitive tasks like HN scraping to DB inserts without token burn.

Verdict

Solid docs and demo video make it approachable, but 11 stars and 1.0% credibility score signal early maturity—pair with stable MCP servers first. Grab it if you're deep in Claude agents; it'll save time on deterministic replays.

(178 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.