dsh-Go
Android client for DeepSeek Harness running local AI agent environments and remote machine control.
Run Linux AI agent environments on Android without root or remotely control your desktop DSH workflows.
dsh-Go is an open-source Android client and execution runtime for DeepSeek Harness. It allows mobile users to execute Linux agent tasks directly on-device without root access or connect remotely to desktop and server instances. It is designed for developers and power users who need mobile control over their AI agents.
What it is
The tool embeds a PRoot-based Ubuntu and Node.js environment into Android to run the DeepSeek Harness engine on-device. It supports model switching, voice input, image attachments, mobile screen automation via accessibility services, and remote management over SSH. It also provides end-to-end encrypted session sharing and customizable UI themes.
Who it's for
This project suits developers, system administrators, and executive assistants who rely on DeepSeek Harness and need to run or monitor agents on Android. It is not intended for casual users looking for a basic chat interface, as it requires managing model API keys and configuring local or remote host settings.
Key features
Local Embedded Linux Runtime
Runs Ubuntu, Node.js, and DSH engines directly on Android using PRoot without needing root access.
Mobile Device Automation
Uses Android accessibility permissions to read screen text, perform clicks, and enter text automatically.
Remote Host Control
Connects to desktop or server-based DeepSeek Harness instances over SSH or the DSH Remote protocol.
Encrypted Session Sharing
Generates end-to-end encrypted web links to share and replay full chat histories in a browser.
Up and running in 3 steps
Our honest verdict
What's great
- Bundles a full PRoot Linux environment to run DSH locally without root privileges.
- Allows offloading agent execution to remote desktop hosts over SSH.
- Includes optional Android screen automation for mobile task execution.
- End-to-end encrypted link creation keeps shared conversation histories private.
Where it falls short
- Local execution requires downloading an embedded Linux environment using around 1 GB of storage.
- Requires sourcing and paying for separate LLM API keys from third-party providers.
- Aggressive Android OEM battery optimizations can terminate background agent tasks.
- Mobile automation features require configuring sensitive accessibility or Shizuku permissions.
Pricing & license
dsh-Go is free to use under the AGPL-3.0 license. Operating costs consist entirely of your third-party LLM API usage fees and any server costs incurred if hosting remote DSH instances.
Open remote AI-training roles
880 open remote AI-training roles, updated November 15, 2026; most pay $20–$175 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
