deepseek-harness-relay-mcp
Run the following command in DeepSeek Harness:
dsh plugin install tonytanglab/deepseek-harness-relay-mcp
Paste the following prompt into your AI chat to install this plugin:
请在 DeepSeek Harness 中运行安装命令以使用插件：dsh plugin install tonytanglab/deepseek-harness-relay-mcp，插件源码位于 https://github.com/tonytanglab/deepseek-harness-relay-mcp。
About this plugin
Harness Relay MCP addresses the limitation of standard MCP tools, which often cannot handle long-running tasks requiring persistent sessions, complex permissions, or deep environment integration. It acts as a bridge between agents and the native DeepSeek Harness, allowing users to delegate complex development workflows without the overhead of command-line operations or manual state management.
Its core capability lies in transforming Harness into a "control plane" managed by MCP. It not only supports asynchronous lifecycle management (start, monitor, recover) but also fully exposes Harness's native plugins, tools, sandbox, and permission models. Whether for read-only access, workspace-write, or full-access permissions, users can seamlessly select these through MCP clients and obtain verifiable web session links, ensuring tasks are traceable and recoverable.
This plugin is tailored for developers seeking efficient agent workflows. It is particularly suitable for modern MCP clients like Codex, Claude Code, and Cursor, enabling these tools to leverage Harness's power rather than just calling model APIs. For tasks involving complex code reviews, implementations, or persistent context, Harness Relay MCP provides a robust and structured solution.
Use Cases
- Delegate complex code implementation tasks.
- Long-running tasks requiring persistent sessions.
- Manage agent workflows via MCP tools.
Best For
- Engineers using MCP agents like Codex or Claude Code.
- Developers needing DeepSeek Harness integration.
- DevOps users managing complex workflows.
Related Plugins
A method pack that makes AI coding agents plan against your real baseline, prove completion with fresh evidence, and reduce reworks and unsafe changes.
Turns the DeepSeek Harness session into a captain that builds a durable sub-agent team, splits goals into dependency-aware tasks, and coordinates work via direct messages and a live Web UI.
Gives coding agents design judgment, letting Claude Code, Cursor, and 70+ agents generate and iterate high-quality UI, presentations, and graphics right from the terminal.
Run the Pi ecosystem's plugins on DeepSeek Harness, unmodified, via a compatibility layer that implements Pi's public extension ABI on DSH's native services.