In DeepSeek Harness (DSH), managing servers typically requires opening a separate SSH terminal. The dsh-vps-manager plugin solves this problem, allowing you to operate VPSs directly from DSH’s chat interface. In addition to curated zero-token commands, the large language model in the conversation can also perform operations on the server.

This is a DSH client plugin maintained by AIcivilization. It connects to VPSs over SSH and provides a complete interactive interface, including a terminal, file browser, and status monitoring, so that server management is integrated into your AI workflow.

Core Features

Zero-Token Commands and VPS Mode
The plugin includes 21 built-in /vps- commands for checking status, installing software, or restarting machines. These operations do not consume model tokens. Click the “VPS Mode” toggle in the chat header to bind the conversation to a specific server. You can then have the AI command the server or operate it yourself through the terminal.

Terminal and Files in the Conversation
The terminal is implemented with ssh -tt and xterm.js, supports interactive programs such as top and vim, and can be minimized/restored. The file browser allows you to drag and drop files for upload, right-click to download, or double-click to edit files. All files are automatically backed up before being overwritten.

Status Monitoring and AI Tools
The plugin provides a server status overview, including items that require attention, such as disk usage, service status, and certificate expiry. Click “Interpret” to get a model-generated summary. AI tools are categorized into three risk levels: read-only (executed automatically), changes (require confirmation), and dangerous operations (require secondary confirmation).

Security and Compatibility
The plugin includes three security mechanisms: backups before file edits, automatic rollback for firewall/SSH changes, and running long-running operations as remote tasks. It supports Windows, macOS, and Linux, and is compatible with both DSH Desktop and the dsh web command-line mode.

Installation and Enablement

Search for dsh-vps-manager in the DSH plugin marketplace and click Install, or run the following command in the terminal (if using dsh web, replace desktop with web):

dsh plugin --profile desktop add dsh-vps-manager

After installation, open DSH and go to Settings → VPS Manager to add a machine.

Typical Usage

  1. Add a machine: Add a machine on the VPS Manager page. Enter the password once; the plugin will automatically configure SSH keys, and the password itself will not be stored. You can also import configuration from ~/.ssh/config.
  2. View status: Click the “Interpret” button in the conversation; the model will generate a brief report based on the server health check results.
  3. Work with files: Right-click a file in the file browser and select “Let AI Check This File”; the model will read the content and provide suggestions.
  4. Use commands: Enter /vps-status or other /vps- commands in the conversation.
  5. Open terminal: Click the terminal button in the conversation, or open the terminal directly from the file browser.

Use Cases and Notes

  • Use cases: Suited for developers, AI agent developers, and operations staff who frequently code locally or review code while needing to quickly deploy or debug on servers. Multiple computers can manage the same server at the same time.
  • Permissions and dependencies: The plugin runs with the permissions of the current DSH process, and operating servers requires the corresponding SSH permissions. It does not include native modules (no need to compile node-pty), and its only dependencies are ws and yaml, with @deepseek-ai/cordis, @deepseek-ai/dsh-tools, and react as peer dependencies.
  • Security mechanisms: Terminal connections are opened only on the local computer by default, and pass DSH login checks, same-origin checks, and plugin token verification. Concurrent modifications on the same machine are queued for execution.

Conclusion

dsh-vps-manager shifts server control from standalone SSH sessions into DSH’s conversational flow. By using curated commands, visualized file operations, and risk-tiered AI tools, it lowers the barrier to server management while backups and validation mechanisms ensure operational reliability.

GitHub repository