Metadata-Version: 2.4
Name: kaisen
Version: 1.0.0
Summary: KAISEN - terminal coding agent that runs commands, reads and writes files in your own shell
Author: spruky
Project-URL: Homepage, https://kaisen.qzz.io
Project-URL: Documentation, https://kaisen.qzz.io/docs
Project-URL: Keys, https://kaisen.qzz.io/api
Keywords: ai,agent,coding-agent,terminal,tui,cli,kaisen
Classifier: Development Status :: 4 - Beta
Classifier: Environment :: Console
Classifier: Intended Audience :: Developers
Classifier: License :: Other/Proprietary License
Classifier: Operating System :: POSIX :: Linux
Classifier: Operating System :: MacOS :: MacOS X
Classifier: Operating System :: Microsoft :: Windows
Classifier: Programming Language :: Go
Classifier: Programming Language :: Python :: 3
Classifier: Topic :: Software Development :: Code Generators
Classifier: Topic :: Utilities
Requires-Python: >=3.8
Description-Content-Type: text/markdown

# KAISEN

A terminal coding agent. It lives in your shell, not in a browser tab: it reads
your files, writes them, runs commands, checks the output, and keeps going until
the task is actually done.

```
pip install kaisen
kaisen
```

## What it is

- A single self-contained binary with a full-screen terminal UI - no runtime,
  no node_modules, no daemon.
- Real execution. The agent runs commands in your working directory and reads
  their true stdout/stderr; it never tells you to go run something yourself.
- Cards for every step: each command, file read and file write is shown, folded,
  with its real output attached.
- A daily token meter, per-turn token counts, and session titles.

## First run

The agent asks for a Kaisen API key. Create one at
[kaisen.qzz.io/api](https://kaisen.qzz.io/api) - it looks like `sk-kaisen-...`.
It is stored in `~/.kaisen/config.json` with `0600` permissions and sent only to
`kaisen.qzz.io`.

## Models

| Model | Use |
| --- | --- |
| `kaisen-x1` | flagship, deepest reasoning (default) |
| `kaisen-x0.5` | fast, lighter and cheaper |

Switch with `/model`. Reasoning depth is `/effort` (`low`, `medium`, `high`,
`max`) - higher effort keeps working through more steps before it answers.

## Commands

`/help` `/model` `/effort` `/clear` `/reset` `/retry` `/run <cmd>` `/cwd`
`/copy` `/keys` `/sid` `/about` `/exit`

`/run` is yours, not the agent's: it executes a command directly and shows the
output without spending a turn.

## Platforms

Prebuilt wheels: Linux (x86-64, arm64), macOS (Intel, Apple silicon), Windows
(x64, arm64). Anywhere else - Termux, Alpine, BSD - `pip install kaisen` fetches
the fallback wheel, which compiles the bundled Go source once on first run into
`~/.kaisen/bin/` and needs a Go 1.22+ toolchain.

Billing, usage and keys: [kaisen.qzz.io](https://kaisen.qzz.io)
