[GH-ISSUE #232] Add Kun integration guide #86

New issue

Open

opened 2026-10-26 09:45:11 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-26 09:45:11 +00:00

Owner

Originally created by @XingYu-Zhong on GitHub (Jun 18, 2026).
Original GitHub issue: https://github.com/deepseek-ai/awesome-deepseek-agent/issues/232

Tool

Kun

Repository: https://github.com/KunAgent/Kun
Product type: cross-platform desktop AI agent workbench with a bundled local kun serve runtime
DeepSeek support: Kun treats DeepSeek as a first-class/default text and reasoning model provider, with README references to deepseek-v4-pro and deepseek-v4-flash for coding, planning, review, long-context sessions, and model routing.

Why it fits this repository

Kun is built around a requirement-first coding workflow: requirement clarification, requirement documents, design/prototype generation, implementation planning, Todo-driven agent coding, and change review in one GUI flow.

It is also a practical DeepSeek agent integration target because it:

uses a local HTTP/SSE runtime boundary via kun serve;
supports workspace file reading/editing, shell tools, approvals, inline diffs, and review panels;
includes Code and Write workspaces, plus IM/automation entry points such as Feishu/Lark and WeChat;
is designed around cost-effective DeepSeek-first agent workflows rather than treating DeepSeek as only a generic optional provider.

Proposed guide scope

A Kun guide could follow the existing installation -> configuration -> first run pattern:

Installation
- Download a release from https://github.com/KunAgent/Kun/releases, or run from source with Node.js 20+.
Configuration
- Select/configure the DeepSeek provider.
- Use current model names: deepseek-v4-pro and/or deepseek-v4-flash.
- Mention DeepSeek V4's 1M context support and document any Kun-specific context/reasoning settings that are exposed in the UI or config.
First run
- Open Code mode, bind a local workspace, start a requirement-first coding task, approve tool calls, and review the generated changes.
- Optionally cover Write mode or remote IM automation as advanced usage.

Expected PR checklist

Add docs/kun.md.
Add docs/kun.zh-CN.md.
Add Kun to README.md in alphabetical order.
Add Kun to README.zh-CN.md in alphabetical order.
Ensure all examples use deepseek-v4-pro / deepseek-v4-flash instead of deprecated model names.
Mention or configure DeepSeek V4 1M context support according to the repo's contribution guide.
Verify any documented Kun config fields actually take effect in the app/runtime.

Originally created by @XingYu-Zhong on GitHub (Jun 18, 2026). Original GitHub issue: https://github.com/deepseek-ai/awesome-deepseek-agent/issues/232 ## Tool **Kun** - Repository: https://github.com/KunAgent/Kun - Product type: cross-platform desktop AI agent workbench with a bundled local `kun serve` runtime - DeepSeek support: Kun treats DeepSeek as a first-class/default text and reasoning model provider, with README references to `deepseek-v4-pro` and `deepseek-v4-flash` for coding, planning, review, long-context sessions, and model routing. ## Why it fits this repository Kun is built around a requirement-first coding workflow: requirement clarification, requirement documents, design/prototype generation, implementation planning, Todo-driven agent coding, and change review in one GUI flow. It is also a practical DeepSeek agent integration target because it: - uses a local HTTP/SSE runtime boundary via `kun serve`; - supports workspace file reading/editing, shell tools, approvals, inline diffs, and review panels; - includes Code and Write workspaces, plus IM/automation entry points such as Feishu/Lark and WeChat; - is designed around cost-effective DeepSeek-first agent workflows rather than treating DeepSeek as only a generic optional provider. ## Proposed guide scope A Kun guide could follow the existing `installation -> configuration -> first run` pattern: 1. Installation - Download a release from https://github.com/KunAgent/Kun/releases, or run from source with Node.js 20+. 2. Configuration - Select/configure the DeepSeek provider. - Use current model names: `deepseek-v4-pro` and/or `deepseek-v4-flash`. - Mention DeepSeek V4's 1M context support and document any Kun-specific context/reasoning settings that are exposed in the UI or config. 3. First run - Open Code mode, bind a local workspace, start a requirement-first coding task, approve tool calls, and review the generated changes. - Optionally cover Write mode or remote IM automation as advanced usage. ## Suggested README entries English table entry: ```md | **Kun** | Cross-platform desktop AI agent workbench for requirement-first coding workflows, with DeepSeek-first model presets, a local `kun serve` runtime, tools, approvals, planning, review, Write mode, and IM automation. | [Guide](./docs/kun.md) | ``` Simplified Chinese table entry: ```md | **Kun** | 面向需求先行 coding 的跨平台桌面 AI Agent 工作台，内置本地 `kun serve` 运行时，优先支持 DeepSeek 模型预设，并覆盖需求、计划、编码、审查、写作和 IM 自动化。 | [指南](./docs/kun.zh-CN.md) | ``` ## Expected PR checklist - [ ] Add `docs/kun.md`. - [ ] Add `docs/kun.zh-CN.md`. - [ ] Add Kun to `README.md` in alphabetical order. - [ ] Add Kun to `README.zh-CN.md` in alphabetical order. - [ ] Ensure all examples use `deepseek-v4-pro` / `deepseek-v4-flash` instead of deprecated model names. - [ ] Mention or configure DeepSeek V4 1M context support according to the repo's contribution guide. - [ ] Verify any documented Kun config fields actually take effect in the app/runtime.

bsakaguchi referenced this issue

2026-10-26 09:46:40 +00:00

[PR #86] docs: fix github copilot deployment method #193