Introduction

In DeepSeek Harness (DSH) use cases, as the number of tools increases, the size and cost of context transfer grow linearly. When connecting to 1,000+ MCP (Model Context Protocol)-compatible tools, traditional data-passing methods often create a huge schema overhead, not only increasing AI bills but also slowing down response times. dsh-mcp-lens is a tool designed for these scenarios. It compresses data structures, consolidates a large tool set into two simple interfaces, and enables AI to call remote APIs efficiently and at low cost.

What Is It?

dsh-mcp-lens is a DeepSeek Harness plugin focused on context optimization and routing for MCP tools. It is maintained by Andrietteprotective835 and released under the MIT License. Its core logic is to compress a massive set of tool instructions into two concise formats, reducing data transfer and supporting progressive disclosure.

Core Features

  1. Data Compression: Reduces data size by up to 90% through context optimization.
  2. Tool Routing: Supports 1,000+ tools but exposes only two interfaces to the model, enabling progressive disclosure.
  3. Performance Improvement: Reduces response time by 50% or more.
  4. Cost Control: Lowers AI bills and supports CI budget Action functionality.
  5. Local Computation: Built-in calculator handles basic arithmetic without calling an AI.
  6. Security Compliance: Processes information on the local machine and follows current security standards.

Installation and Enablement

The plugin runs as a local program. No complex package management command is required; download it and start it directly.

  1. Download: Click the green button at the top of the page or the link in the download section to obtain the archive.
  2. Run: Save the file locally and run the program to start using it immediately.
  3. Environment Requirements:
    • Operating system: Windows 10 or Windows 11.
    • Node.js: version ^22.19.0 or >=24.0.0.
  4. Dependencies: Make sure the following peer dependencies are installed:
    {
      "@deepseek-ai/cordis": "^4.0.1",
      "@deepseek-ai/dsh-subprocess": "^0.1.0-rc.6",
      "@deepseek-ai/dsh-tools": "^0.1.0-rc.6"
    }

Typical Usage

dsh-mcp-lens works by organizing AI instructions into two clean paths. It compresses complex tool definitions so DeepSeek Harness processes only the necessary context. This enables AI to search for and call remote APIs quickly without being slowed down by a huge context window.

Use Cases and Notes

  • Use Cases: Developers who need to manage a large number of MCP tools, want to reduce API call costs, and care about response speed.
  • Privacy and Security: All information is processed on the local machine and sensitive data is not transferred externally.
  • Permission Note: The plugin runs with the current DSH process permissions. Please make sure to review the source code and license before installation.

Conclusion

For DeepSeek Harness users who need to manage the MCP ecosystem efficiently, dsh-mcp-lens provides a complete solution from context compression to cost control.

  • GitHub: https://github.com/Andrietteprotective835/dsh-mcp-lens
  • Ecosystem Directory: https://www.skillhub.cn/plugins/Andrietteprotective835/dsh-mcp-lens