Metadata-Version: 2.5
Name: kin
Version: 0.28.0
Summary: Native Python agent harness with a Textual terminal UI
Project-URL: Documentation, https://kinra.ai/docs/kin/
Project-URL: Repository, https://github.com/kinra-ai/kin
Project-URL: Issues, https://github.com/kinra-ai/kin/issues
License-Expression: Apache-2.0
License-File: LICENSE
Requires-Python: >=3.11
Requires-Dist: anthropic<2,>=1
Requires-Dist: anyio<5,>=4.5
Requires-Dist: defusedxml
Requires-Dist: httpx
Requires-Dist: jsonschema
Requires-Dist: mcp<3,>=2
Requires-Dist: nbformat>=5.9
Requires-Dist: openai<3,>=2.43
Requires-Dist: openpyxl
Requires-Dist: pathspec
Requires-Dist: pygments
Requires-Dist: pypdf>=6.16.1
Requires-Dist: python-docx
Requires-Dist: rapidfuzz
Requires-Dist: regex
Requires-Dist: rich>=13
Requires-Dist: textual-speedups>=0.2.1
Requires-Dist: textual>=8.2.8
Requires-Dist: tomlkit
Requires-Dist: trafilatura<2.2,>=2.1
Requires-Dist: tree-sitter-markdown>=0.3
Requires-Dist: tree-sitter>=0.25
Provides-Extra: site-drive
Requires-Dist: playwright>=1.61; extra == 'site-drive'
Description-Content-Type: text/markdown

<h1 align="center">
  <img src="docs/assets/kinra-mark.svg" alt="" width="30" height="30">
  <code>Kin</code>
</h1>

<p align="center">
  <em>A terminal-native AI peer that works with you in a real filesystem.</em>
</p>

**Kin** reads files, changes code, runs tools, and researches alongside you.
Start it in a directory and work with the files already there. Conversations
persist, tools expose their results, and approvals keep authority explicit.

[![Kin welcome screen with the compact emblem, active model, workspace, trust posture, and message composer.](docs/assets/screenshots/welcome_banner.svg)](docs/assets/screenshots/welcome_banner.svg)

## Install and start

On Linux or Apple Silicon macOS:

```bash
bash <(curl -fsSL https://get.kinra.ai/install.sh)
kin doctor
kin
```

Use `/providers` to connect a model, then describe what you want to do.
The [installation guide](https://kinra.ai/docs/kin/getting-started/install/)
covers WSL2, updates, verification, and rollback. Intel macOS is best-effort
and unqualified because its native dependency stack is upstream-unsupported.

Continue with [your first session](https://kinra.ai/docs/kin/getting-started/first-session/)
or the [full manual](https://kinra.ai/docs/kin/).

## Work with Kin

- **Use your workspace.** Read and edit files, run shell commands, search
  locally, retrieve web sources, and connect MCP services. See
  [Files and tools](https://kinra.ai/docs/kin/guide/tools/).
- **Keep context across sessions.** Resume conversations and maintain visible
  workspace instructions. Separate conversations still share files; use Git
  worktrees when edits need isolation. See
  [Sessions](https://kinra.ai/docs/kin/guide/sessions/).
- **Delegate and reuse procedures.** Give
  [subagents](https://kinra.ai/docs/kin/guide/subagents/) bounded work and use
  [skills](https://kinra.ai/docs/kin/guide/skills/) for repeatable instructions.
- **Control local authority.** Workspace trust, `auto` and `strict` modes,
  planning freeze, and protected file-tool writes govern execution. Shell and
  code run with your local authority once permitted. See
  [Modes and approvals](https://kinra.ai/docs/kin/guide/modes-and-permissions/).
- **Choose how sessions run.** Use the terminal, an optional Herdr host for
  detach/reattach, or a [local client](https://kinra.ai/docs/kin/guide/local-clients/).
  Host schedulers can invoke `kin -p` for
  [unattended work](https://kinra.ai/docs/kin/guide/scheduled-jobs/).

[![Kin timeline with reasoning, shell output, edit details, and interruption states.](docs/assets/screenshots/timeline_states.svg)](docs/assets/screenshots/timeline_states.svg)

## The peer stance

Calling Kin a peer is an engineering stance: give it clear intent, useful
context, inspectable tools, and room to question assumptions and recover.
Claims still need evidence, and completed work still needs verification.
Read the [AI peer philosophy](https://kinra.ai/docs/kin/concepts/ai-peer-philosophy/)
and [product direction](docs/product-direction.md) for the principles and
deliberate limits behind the product.

## Develop

Kin is a native Python harness with a Textual client and a typed event boundary.
Install `uv` and Task v3, then use the pinned checkout environment:

```bash
task setup
task run
```

[CONTRIBUTING.md](CONTRIBUTING.md) covers changes to `dev`, verification, and
the public-manual handoff. [AGENTS.md](AGENTS.md) is the AI peer's working map;
`CLAUDE.md` includes it. [Maintainer guides](docs/internals/README.md) cover
implementation and operations. [STATUS.md](STATUS.md) records current work.

`task ship` is the local promotion proof; there is no hosted CI promotion gate.
`main` contains verified releases. The [release guide](docs/internals/release.md)
owns publication and recovery.

The public manual and [changelog](https://kinra.ai/docs/kin/changelog/) are
authored in `kinra-site`. Contributors without access can propose wording as
described in [Contributing](CONTRIBUTING.md#public-manual-changes).

## License and stewardship

Kin is [Apache-2.0 licensed](LICENSE) and maintainer-led. Contributions are
welcome; the maintainer retains scope and release authority, with no obligation
to respond, merge, implement, maintain, or support. See
[CONTRIBUTING.md](CONTRIBUTING.md) for terms,
[SECURITY.md](SECURITY.md) for private reports,
[TRADEMARKS.md](TRADEMARKS.md) for name and mark restrictions, and
[THIRD_PARTY_NOTICES.md](THIRD_PARTY_NOTICES.md) for upstream notices.
