DeepSeek Harness VPS
About
DeepSeek Harness (the command is dsh) is DeepSeek's open-source agent harness. It runs AI agents that call models over an API, and it is built on the idea that everything is a plugin.
The installed version is 0.1.0rc7, a Developer Preview release. Features and usage can still change, so it suits trying things out and building, not systems that must stay fixed.
The AI model does not run on this VPS. The VPS runs the harness and its workspace, and the model is called over an API. You use your own DeepSeek API key, or another supported provider.
How it works
DeepSeek Harness needs 4 GB RAM, so only plans with 4 GB or more are offered. Regions: Jakarta, Singapore, Bangkok, Tokyo, Seoul or Johor Bahru.
The VPS is created with DeepSeek Harness 0.1.0rc7 for the ubuntu user. The web workspace runs on 127.0.0.1:8080 and is deliberately not exposed to the internet.
SSH in as ubuntu and run cat ~/.dsh/.credentials.yaml to see the workspace login. From your computer, run ssh -L 8080:127.0.0.1:8080 ubuntu@IP.
Browse to http://localhost:8080 while the tunnel is open, log in, then enter your DeepSeek or other provider API key.
Pricing
Ideal for an Indonesian audience
Ideal for an Asian audience
Ideal for an audience in Malaysia and nearby
Ideal for an audience in Thailand and nearby
Ideal for an audience in Japan and nearby
Ideal for an audience in South Korea and nearby
Comparison
| Your own laptop or PC | Helipod DeepSeek Harness VPS | |
|---|---|---|
| Agent keeps running when the laptop is off | ||
| Installing dsh | Install it yourself | Installed when the VPS is ready |
| Separate from your personal files | ||
| RAM for the harness | Shared with your other apps | 4 GB or more for the server alone |
| Workspace access | Directly on localhost | Through an SSH tunnel to localhost |
| Cost | No server cost | From Rp {price}/month |
Features
Use cases
FAQ