Metadata-Version: 2.4
Name: kanonry
Version: 0.0.1
Summary: Record, replay, and verify AI agent runs. Coming soon.
Author-email: Abhiyash <abhiyashnibber@gmail.com>
License-Expression: MIT
Project-URL: Homepage, https://kanonry.com
Project-URL: Repository, https://github.com/kanonry/kanonry
Requires-Python: >=3.9
Description-Content-Type: text/markdown
License-File: LICENSE
Dynamic: license-file

# Kanonry

Record, replay, and verify AI agent runs.

Kanonry is building developer-first verification tooling for AI agents.

This package is a placeholder release used to reserve the `kanonry` package name on PyPI while the first public version is in development.

Visit [kanonry.com](https://kanonry.com) for updates and to join the waitlist.

## What is Kanonry?

Kanonry is a verification layer for AI agents.

It is being built to help developers capture real agent executions, replay them deterministically, compare outcomes across runs, and turn assertions over recorded behavior into checks that can run locally and in CI.

The goal is to make agent behavior easier to inspect, test, and trust before it reaches production.

## Planned capabilities

Kanonry is intended to support workflows like:

- recording real agent runs
- replaying runs with stable inputs
- diffing outputs and tool behavior across runs
- writing assertions over recorded behavior
- running verification checks in local development and CI

## Status

This is a pre-release placeholder.

There is no public API yet, and the package does not currently provide production functionality.

## Installation

```bash
pip install kanonry
