Deepseek-Harness VSCode Integration Community Edition

Please note that this README is maintained and translated by chatgpt models,it's better to refer to README.zh-CN.md for human-friendly descripitons.

Feel free to leave issues!

A full-featured community VS Code extension for connecting to DeepSeek Harness and completing the Agent workflow without leaving the editor.

English | 简体中文

Note: This is an independent community project. It is not an official DeepSeek project and is not maintained by DeepSeek.

[!NOTE] Highlights

• Stream tasks in persistent Harness sessions, with queueing, steering, approvals and plan reviews
• Type @ to search workspace files in real time and insert an explicit file reference
• Attach selections, diagnostics and Git diffs with byte-size and truncation feedback before sending
• Inspect tool calls, results and reasoning in a built-in Trace, with clickable file and line locations
• Review per-turn file changes with native VS Code diffs before restoring anything
• Monitor the active model, reasoning effort, billed tokens, cache usage and estimated context pressure

The extension is designed as a complete working surface rather than a thin chat wrapper: session state, IDE context, runtime activity, approvals, traces and file changes stay connected throughout a task.

Architecture

The chat interface is a React Webview backed by a typed, full-state bridge. The Extension Host is responsible for VS Code APIs, Runtime RPC, credentials, secure Markdown rendering and action validation.

Installation

From the Extension Marketplace

🔗 Install from the Marketplace

From GitHub Releases

Download the .vsix package from GitHub Releases, then run Extensions: Install from VSIX....

Build from source

npm install
npm run check
npm run package

Install the generated .vsix via Extensions: Install from VSIX....

Usage

1. Open a trusted workspace.
2. Open DSH Chat (Ctrl+Shift+Alt+D / Cmd+Shift+Alt+D).
3. Type @ to search and reference workspace files, attach selections, diagnostics or unstaged Git diffs, or type / to open the local command menu.
4. Send a prompt and handle tool approvals, questions and plan reviews directly in the chat view.

The Explorer context menu includes DSH: Ask About This Resource, which pre-fills the selected file or directory and its workspace root without sending automatically. Use DSH: Diagnose Environment to write a redacted runtime and command-discovery report to the DSH output channel.

If dsh reports a missing or invalid API key, click Key in the chat header or run DSH: Configure API Key. The key is passed to dsh's credential service, and an encrypted copy is stored in VS Code SecretStorage for the balance indicator. It is not written to prompts, extension state or logs.

Configuration

{
  "dsh.command": "npx",
  "dsh.commandArgs": ["-y", "@deepseek-ai/dsh", "web"]
}

• dsh.command / dsh.commandArgs: command used to start dsh web.
• dsh.serverUrl: connect to an existing runtime instead of starting a new process.
• dsh.serverPort: local runtime port; 0 means auto-select.
• dsh.maxContextBytes: maximum UTF-8 byte size of the <ide_context> block.
• dsh.apiKeyEnv: credential reference used by DSH: Configure API Key (defaults to DEEPSEEK_API_KEY).
• dsh.balanceRefreshIntervalMs: DeepSeek balance refresh interval (defaults to 30 seconds).

The balance indicator calls DeepSeek's official /user/balance endpoint. Chat credentials remain managed by the dsh runtime; the encrypted SecretStorage copy is used only for this read-only balance request.

Development

npm install
npm run check      # TypeScript check
npm run compile    # Build to dist/
npm run package    # Compile + vsce package

Press F5 in VS Code to launch the Extension Development Host.

More Information

• Changelog
• Product TODO
• Third-party notices

License

MIT