roach-code 是 AI Skill Hub 本期精选Agent工作流之一。综合评分 7.5 分,整体质量较高。我们推荐使用将其纳入你的 AI 工具库,帮助提升工作效率。
roach-code 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排,将复杂的多步骤任务拆解为清晰的自动化流程,实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成,适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
roach-code 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排,将复杂的多步骤任务拆解为清晰的自动化流程,实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成,适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
# 方式一:go install(推荐) go install github.com/tmdgusya/roach-code@latest # 方式二:从源码编译 git clone https://github.com/tmdgusya/roach-code cd roach-code go build -o roach-code . # 方式三:下载预编译二进制 # 访问 Releases 页面下载对应平台二进制文件 # https://github.com/tmdgusya/roach-code/releases
# 查看帮助 roach-code --help # 基本运行 roach-code [options] <input> # 详细使用说明请查阅文档 # https://github.com/tmdgusya/roach-code
# roach-code 配置说明 # 查看配置选项 roach-code --config-example > config.yml # 常见配置项 # output_dir: ./output # log_level: info # workers: 4 # 环境变量(覆盖配置文件) export ROACH_CODE_CONFIG="/path/to/config.yml"
<p align="center"> <img src="docs/mascot.png" alt="Roach Code mascot" width="200"/> </p> <p align="center"> <img src="docs/logo.svg" alt="Roach Code" width="440"/> </p>
<p align="center"> <strong>English</strong> · <a href="./README.zh-CN.md">简体中文</a> · <a href="./README.ko.md">한국어</a> · <a href="./docs/SPEC.md">Spec</a> </p>
<p align="center"> <a href="https://github.com/tmdgusya/roach-code/actions/workflows/ci.yml"><img src="https://img.shields.io/github/actions/workflow/status/tmdgusya/roach-code/ci.yml?style=flat-square&label=ci&labelColor=161b22&logo=githubactions&logoColor=white" alt="CI"/></a> <a href="./LICENSE"><img src="https://img.shields.io/github/license/tmdgusya/roach-code?style=flat-square&color=8b949e&labelColor=161b22" alt="license"/></a> <a href="https://github.com/tmdgusya/roach-code/stargazers"><img src="https://img.shields.io/github/stars/tmdgusya/roach-code.svg?style=flat-square&color=dbab09&labelColor=161b22&logo=github&logoColor=white" alt="GitHub stars"/></a> <a href="https://github.com/tmdgusya/roach-code/graphs/contributors"><img src="https://img.shields.io/github/contributors/tmdgusya/roach-code.svg?style=flat-square&color=bc8cff&labelColor=161b22&logo=github&logoColor=white" alt="contributors"/></a> <a href="https://github.com/tmdgusya/roach-code/discussions"><img src="https://img.shields.io/github/discussions/tmdgusya/roach-code.svg?style=flat-square&color=58a6ff&labelColor=161b22&logo=github&logoColor=white" alt="Discussions"/></a> </p>
<br/>
A config- and plugin-driven harness — a single static Go binary, tuned around prefix caching so token costs stay low across long sessions.
<br/>
Roach Code is a multi-model rewrite of deepseek-reasonix (by @esengine). It keeps the Reasonix harness and generalizes it from DeepSeek-only to any provider — Codex/OpenAI (Responses API + ChatGPT OAuth), MiniMax, GLM (Z.ai), Anthropic, and any OpenAI-compatible endpoint. Not a from-scratch project: a rebrand + multi-provider extension of upstream's work.
<br/>
- Config-driven. Providers, the agent, enabled tools, and plugins are all declared in roach-code.toml. No hardcoded models. - Multi-model & composable. Any OpenAI-compatible endpoint is a config entry, not new code. Optionally run two models together (executor + planner) in separate, cache-stable sessions. - Plugin-driven. External tools run as subprocesses over stdio JSON-RPC (MCP-compatible). Built-in tools self-register at compile time. - Zero-friction distribution. CGO_ENABLED=0 single binary; cross-compile to six targets with one command. The only dependency is a TOML parser.
Prebuilt binary (no Go toolchain required) — installs from the latest GitHub release:
```sh
make build # -> bin/roach-code
make cross # -> dist/ (darwin|linux|windows × amd64|arm64)
roach-code setup # config wizard → ./roach-code.toml
export DEEPSEEK_API_KEY=sk-... # or put it in ~/.env (see .env.example)
roach-code chat # then run /init to generate AGENTS.md (project memory)
roach-code run "implement the TODOs in main.go"
roach-code run --model mimo-pro "add unit tests for this function"
echo "explain this code" | roach-code run
Installed binaries also answer to the short alias roach (e.g. roach chat). A few more commands:
roach-code models # list configured providers / models
roach-code models refresh # re-fetch each provider's model list from its /models API
roach-code codex login # sign in to Codex with a ChatGPT subscription (OAuth)
roach-code update # self-update to the latest release
Resolution order: flag > ./roach-code.toml > ~/.config/roach-code/config.toml > built-in defaults. Secrets come from the environment via api_key_env and are never stored in config files.
```toml default_model = "deepseek-flash" # executor; set [agent].planner_model to add a planner
roach-code setup keeps first-run minimal: pick provider → keys (every SKU of a chosen provider is enabled). Running two models together (executor + planner, separate cache-stable sessions) is a one-line edit afterwards — set planner_model to any other enabled provider:
[agent]
planner_model = "deepseek-pro" # used as the low-frequency planner
Subagent skills inherit the executor model by default. Set subagent_model to run them on another configured model, or use subagent_models to override only specific skills such as review or security_review.
Embed @ references in a message and Roach Code resolves them before sending, as tagged context blocks: @path/to/file (or @dir) injects a local file's contents (or a directory listing), and @<server>:<uri> injects an MCP resource. A local path is only treated as a reference when it actually exists, so ordinary @mentions stay literal. Typing / or @ opens an autocomplete menu — slash commands, or hierarchical file navigation (one directory level at a time, descend into folders) plus MCP resources.
Roach Code is an MCP client. A [[plugins]] entry's type selects the transport: stdio (default) launches a local subprocess (command/args/env); http (Streamable HTTP) connects to a remote url with optional static headers (${VAR} / ${VAR:-default} expanded from the environment, so tokens stay out of the file). Tools surface to the model as mcp__<server>__<tool>; a tool declaring MCP's readOnlyHint: true joins parallel dispatch and the permission reader-default.
A server's prompts surface as /mcp__<server>__<prompt> slash commands (positional args after the command); its resources are pulled in by writing @<server>:<uri> in a message; /mcp lists connected servers and what each exposes. make build also produces bin/roach-code-plugin-example — a runnable reference stdio server (echo, wordcount, a review prompt, a style-guide resource) you can copy.
[[plugins]] # local stdio server
name = "example"
command = "roach-code-plugin-example"
[[plugins]] # remote server over Streamable HTTP
name = "stripe"
type = "http"
url = "https://mcp.stripe.com"
headers = { Authorization = "Bearer ${STRIPE_KEY}" }
Already have an .mcp.json? Drop it in the project root and Roach Code reads it as-is — the mcpServers spec (command/args/env, type/url/ headers, ${VAR} expansion) maps field-for-field onto [[plugins]]. Both sources are merged; on a name collision roach-code.toml wins.
{
"mcpServers": {
"filesystem": { "command": "npx", "args": ["-y", "@modelcontextprotocol/server-filesystem", "/path"] },
"stripe": { "type": "http", "url": "https://mcp.stripe.com", "headers": { "Authorization": "Bearer ${STRIPE_KEY}" } }
}
}
中文项目简介
Roach Code 支持配置驱动、多模型和可组合的特性。任何 OpenAI 兼容的端点都可以通过配置文件来声明,而不需要编写新代码。Roach Code 还支持插件驱动,外部工具可以作为子进程通过 stdio JSON-RPC 运行。
环境依赖与系统要求中文说明
Roach Code 可以通过预编译二进制文件(无需 Go 工具链)或从源代码编译来安装。预编译二进制文件从最新的 GitHub 发布版本安装: ```sh # 预编译二进制文件(无需 Go 工具链) # 从最新的 GitHub 发布版本安装 # 从源代码编译 make build # -> bin/roach-code make cross # -> dist/ (darwin|linux|windows × amd64|arm64) ```
快速入门 ```sh roach-code setup # config wizard → ./roach-code.toml export DEEPSEEK_API_KEY=sk-... # 或者将其放入 ~/.env (参见 .env.example) roach-code chat # 然后运行 /init 来生成 AGENTS.md (项目内存) roach-code run "implement the TODOs in main.go" roach-code run --model mimo-pro "add unit tests for this function" echo "explain this code" | roach-code run ```
配置说明 配置解析顺序:**flag > ./roach-code.toml > ~/.config/roach-code/config.toml > 内置默认值**。密钥来自环境变量通过 api_key_env,并且永远不会存储在配置文件中。 ```toml default_model = "deepseek-flash" # executor; set [agent].planner_model 来添加一个规划器 planner_model = "mimo-pro" # 可选低频率 subagent_model = "deepseek-pro" # 可选默认值 ```
API/接口说明
Roach Code 是一个 MCP 客户端。[[plugins]] 条目中的 type 选择传输:stdio (默认) 启动一个本地子进程(command / args / env);http (Streamable HTTP) 连接到一个远程 url(url)并且可选的静态 headers(${VAR} / ${VAR:-default})从环境中扩展。 ```toml [[plugins]] type = "stdio" cmd = "roach-plugin" [[plugins]] type = "http" url = "https://example.com" headers = {\n \"Authorization\" = \"Bearer ${TOKEN}\"\n} ```
roach-code是一个有趣的AI工作流项目,值得关注
AI Skill Hub 为第三方内容聚合平台,本页面信息基于公开数据整理,不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后,再部署至生产环境,并做好必要的安全评估。
✅ MIT 协议 — 最宽松的开源协议之一,可自由商用、修改、分发,仅需保留版权声明。
经综合评估,roach-code 在Agent工作流赛道中表现稳健,质量良好。如果你已有明确的使用需求,可以直接上手体验;如果还在评估阶段,建议对比同类工具后再做决策。
| 原始名称 | roach-code |
| 原始描述 | 开源AI工作流:A config- and plugin-driven terminal AI coding agent (multi-model: DeepSeek, Cod。⭐22 · Go |
| Topics | AI工作流Go |
| GitHub | https://github.com/tmdgusya/roach-code |
| License | MIT |
| 语言 | Go |
收录时间:2026-06-07 · 更新时间:2026-06-08 · License:MIT · AI Skill Hub 不对第三方内容的准确性作法律背书。
选择 Agent 类型,复制安装指令后粘贴到对应客户端