CLI Alpha • Core MIT • UI BSL-1.1

DeepAgent — 为 DeepSeek
定制的 AI Agent

一条命令安装，开箱即用。不是 IDE 里的编程助手，也不是包装单一 API 的聊天工具。 这是一个AI Native Agent，运行在你的服务器上，记住它学到的东西，运行越久越强大。

$ curl -fsSL https://deepseekagent.starseas.org/install.sh | bash

CLI Alpha supports macOS Apple Silicon only · No sudo

查看 CLI Alpha Release 查看安装与卸载说明

Get started in 60 seconds

1

Install

curl -fsSL https://deepseekagent.starseas.org/install.sh | bash

Installs a checksum-verified, versioned Core release under ~/.deepagent. No sudo needed.

2

Configure

bash 
# Interactive setup wizard
deepagent setup

# Or choose your model
deepagent model

连接 DeepSeek API、OpenRouter，或自定义端点。

3

Start chatting

bash 
deepagent

That's it. Interactive CLI with tools, local sessions, and memory.

4

Verify the installation

bash 
# Check configuration and runtime health
deepagent doctor

Resolve any provider or local environment problem before the first real task.

5

Keep it up to date

bash 
deepagent update

Reads the promoted Alpha channel, verifies the immutable release manifest and SHA-256, then switches versions.

本轮不承诺 Intel macOS、Linux、Windows、WSL 或 Android。 安装器会在不支持的平台明确停止，不会尝试实验性安装。

See it in action

deepagent

Features

为 DeepSeek 深度定制

针对 DeepSeek 模型深度优化，充分发挥 DeepSeek 在推理和代码生成方面的能力。

AI Native Agent

不是 IDE 插件，不是聊天机器人封装。是一个自主运行的 Agent，记住它学到的东西，运行越久越强大。

独立安装与数据目录

所有产品文件位于 ~/.deepagent，不读取或覆盖 Hermes 与用户 OpenCode 数据。

可验证更新与回滚

版本 Manifest、文件大小和 SHA-256 全部匹配后才切换版本，失败保留当前版本。

安全卸载

卸载只处理安装 Manifest 登记且通过路径边界校验的文件，可选择保留用户数据。

AI Native Agent 架构

深度融合 AI Agent 与研发工具链，内置虚拟研发团队，开箱即用。

Tools

40+ built-in — web search, terminal, file system, browser automation, vision, image generation, text-to-speech, code execution, subagent delegation, memory, task planning, cron scheduling, multi-model reasoning, and more.

Supported release

CLI Alpha is tested and released only for macOS Apple Silicon. Connect to DeepSeek API, OpenRouter, or a configured OpenAI-compatible API.

Environments

Run locally, in Docker, over SSH, on Modal, Daytona, or Singularity. Container hardening with read-only root, dropped capabilities, and namespace isolation.

Skills

40+ bundled skills covering MLOps, GitHub workflows, research, and more. The agent creates new skills on the fly and shares them via the open agentskills.io format. Install community skills from ClawHub, LobeHub, and GitHub.

Research

Batch trajectory generation with parallel workers and checkpointing. Atropos integration for RL training. Export to ShareGPT for fine-tuning with trajectory compression.