---
title: The Agent Reliability Stack
emoji: 🪡
colorFrom: yellow
colorTo: gray
sdk: gradio
sdk_version: "5.49.1"
python_version: "3.12"
app_file: app.py
pinned: false
license: mit
short_description: "Live demo of fit, guard, snap, vet, cast — five small libs."
tags:
  - llm
  - agent
  - reliability
  - tool-use
  - structured-output
  - prompt-injection
---

# The Agent Reliability Stack — Live Demo

Five small, focused libraries that fix the boring problems every long-running AI agent eventually hits. Pure Python, zero runtime dependencies. BYO LLM.

Pick a tab to see what each library does against your own input — no install needed.

## What you can try

- **🪟 fit** — fit a chat history into a token budget (drop-oldest / drop-middle / priority)
- **🛡️ guard** — check URLs against a network-egress allowlist before any tool fetch
- **📸 snap** — diff two tool-call traces, catch silent regressions
- **✅ vet** — validate tool-call args before execution; LLM-friendly retry hints
- **🎯 cast** — extract JSON from messy LLM text and validate it against a shape

## Install

Pick the language you live in:

```bash
# Python
pip install mk-agentkit         # all five
pip install agentfit-py         # individual

# JavaScript / TypeScript
npm install @mukundakatta/agentkit   # all five
npm install @mukundakatta/agentfit   # individual
```

Or run them as MCP servers inside Claude Desktop / Cursor / Cline / Windsurf / Zed:

```bash
npx -y @mukundakatta/agentfit-mcp
```

## Links

- 🌐 Landing: <https://mukundakatta.github.io/agent-stack/>
- 📦 PyPI: [mukundakatta](https://pypi.org/user/mukundakatta/)
- 📦 npm: [@mukundakatta](https://www.npmjs.com/~mukundakatta)
- 💻 GitHub: [MukundaKatta](https://github.com/MukundaKatta)

## License

MIT.