# Attentry

> Attentry is an open-source AI tool for code assistant. Static analyzer that checks AI coding agents' code for security and crash issues, offline. It is available on vscode, CLI and self-hosted.

- **Kind:** Tool
- **Pricing:** Open source
- **Platforms:** vscode, CLI, self-hosted
- **Repository:** https://github.com/aarv-1o/attentry
- **Categories:** [code assistant](https://ai-powered.com/categories/code-assistant)
- **Listing updated:** 2026-11-18

## About

The README says Attentry reads code as an agent writes it, rather than relying on another model. It is a standalone static analyzer that runs locally without an API key, and it reports findings directly to Claude Code and Codex through hooks and MCP tools. When an edit introduces a security or crash risk, the agent is told straight away so it can fix the problem before continuing.

After a task finishes, a change card lists the files touched, the problems added or fixed, and whether the repo's tests passed. Problems that existed before the task are left out, so only new issues surface. Findings also appear as editor underlines in VS Code, Cursor, Windsurf and Antigravity, with explanations drawn from the code itself.

It suits developers who let coding agents write substantial amounts of JavaScript, TypeScript, Python or Astro and want a review step that doesn't depend on AI credits. Block mode, which refuses exact duplicate functions and stops tasks that add security or crash problems, is currently limited to Claude Code.

## Features

- **Live agent feedback:** Tells the coding agent about security and crash risks introduced by each edit so they can be fixed immediately.
- **Offline static analysis:** Runs its own analyzer on your machine with no API key, and the code never leaves the machine.
- **Duplicate and dead code detection:** Finds exact and near copies of functions, unused packages, unused files and uncalled functions.
- **Task change card:** Summarises changed files, new or fixed problems and test results at the end of each task.
- **Editor underlines and fixes:** Shows findings as editor diagnostics with explanations and a light-bulb action to remove unused code.
- **Suggest or block modes:** Choose between advisory suggestions and blocking exact copies or tasks with new security or crash problems.

## Good for

- Catching unauthenticated routes or injection risks in code an AI agent just wrote
- Stopping an agent from re-implementing a helper the repo already has
- Reviewing an agent's changes for crash risks such as unguarded JSON.parse
- Finding unused packages, files and functions after a large generated change
- Checking the frontend and backend API calls match existing routes

## FAQ

### Does Attentry need an API key or use an AI model?

No. The README says it is its own static analyzer, needs no API key and runs on your machine.

### Which agents and editors does it work with?

The README names Claude Code and Codex, and the VS Code extension works in VS Code, Cursor, Windsurf and Antigravity. Other editors can run attentry lsp as a language server.

### Which languages are supported?

JavaScript, TypeScript (including JSX and TSX), Python and Astro.

### Is block mode available for Codex?

Not yet. The README says block mode is tested with Claude Code only and Codex block mode comes in a later version.

### Is Attentry free?

Yes. Attentry is open source, so you can use and self-host it for free.

### What are the best alternatives to Attentry?

The closest alternatives to Attentry are Kai, CodeComplete and codesnippets.

## Where this comes from

- Description: written from the project's README.

## Alternatives

- [Kai](https://ai-powered.com/tools/kai-3): Kai is a freemium AI tool for code assistant.
- [CodeComplete](https://ai-powered.com/tools/codecomplete): CodeComplete is an AI tool for code assistant.
- [codesnippets](https://ai-powered.com/tools/codesnippets): codesnippets is a freemium AI tool for generative code.

---
Source: https://ai-powered.com/tools/attentry
