Introduction

The core design philosophy of DeepSeek Harness (DSH) is a plugin-based approach. When building agents or performing tasks such as code review, it is often necessary to connect a local knowledge graph or code memory service to the DSH context. The dsh-codebase-memory-mcp plugin addresses this need. It uses the built-in @deepseek-ai/dsh-mcp-client bridge layer in DSH to connect a local codebase-memory-mcp knowledge graph server directly to the DSH environment, without developers needing to implement the MCP transport layer themselves.

Plugin Overview

This is a DeepSeek Harness plugin designed to bridge a local codebase-memory-mcp knowledge graph MCP server into DSH. By reusing the built-in @deepseek-ai/dsh-mcp-client in DSH, it allows the model to directly call tools related to codebase memory.

Basic Information
* Project Name: dsh-codebase-memory-mcp
* Author: princeofdream
* Category: Memory
* License: MIT

Core Features

This plugin provides the following core capabilities:

  1. Server Bridging: Connects a local codebase-memory-mcp knowledge graph service to DSH.
  2. Tool Registration:
    • MCP Tools: Registers 14 MCP tools prefixed with mcp__codebase-memory__* (such as index_repository, search_graph, trace_path, query_graph, detect_changes, etc.) for the model to call.
    • Host Tools: Registers two host-level tools for monitoring and restarting the service:
      • codebase_memory_status: Queries the service status.
      • codebase_memory_restart: Restarts the service.
  3. Configuration Management: Provides a codebase-memory-mcp namespace and supports configuring the server executable path, working directory, extra arguments, environment variables, and an enable switch.

Installation and Enablement

Install the plugin via the command line:

dsh plugin --profile web add dsh-codebase-memory-mcp

After installation, you need to reload the DSH Web host (restart the application or reload it within the application) to activate the plugin. If you are in a local development environment, you can also use the link: protocol to specify a local path for installation.

Configuration and Usage

Check Service Status

After installation, first confirm whether the bridge is working properly. Use the model to call the host tool codebase_memory_status to check the status.

codebase_memory_status

Fix Configuration and Restart

If the status shows error, it is usually caused by a misconfigured executable file path or an environment issue. The plugin does not crash when the binary is invalid; instead, it logs the issue.

Please follow these steps to fix it:

  1. Open the DSH settings document and find the codebase-memory-mcp namespace.
  2. Check and correct the exePath, cwd, or env configuration items.
  3. Save the settings (hot reloading is supported, so DSH does not need to be restarted).
  4. Call codebase_memory_restart again to restart the bridge.

Verify the Connection

When codebase_memory_status returns state: "ready" and toolCount > 0, the bridge has been established, and the model can start using the mcp__codebase-memory__* tools.

Notes

  1. Dependency Requirement: This plugin depends on the @deepseek-ai/dsh-mcp-client package built into DSH. If installation fails, check whether this dependency has been properly included.
  2. GUI Card Limitation: The GUI card in the configuration interface is a best-effort feature and may not appear in all DSH build versions. If you cannot find this card in settings, you can still manage the codebase-memory-mcp namespace through the generic settings document editor or by directly modifying the configuration file.
  3. Error Handling: If the binary is unavailable, the plugin does not crash. It logs the error to the console, allowing you to continue trying to fix the issue through the host tools.
  4. Startup Status Monitoring: You can observe the startup status through the console plugin loading report or the codebase_memory_status tool.

Summary

dsh-codebase-memory-mcp provides DSH with the ability to connect to a local knowledge graph. With its built-in bridging logic and configuration management, developers can easily integrate the code memory service into their workflows. Before use, it is recommended to check the source code and license, and ensure that all dependencies are satisfied.

  • GitHub Repository: https://github.com/princeofdream/dsh-codebase-memory-mcp
  • Community Directory: https://www.skillhub.cn/plugins/princeofdream/dsh-codebase-memory-mcp