AI Agent Hub
Back to plugins
🤖

dsh-plugin-codegraph

Model Inference Updated 2026.08.25

Run the following command in DeepSeek Harness:

dsh plugin install CC19990113/dsh-plugin-codegraph

Paste the following prompt into your AI chat to install this plugin:

You can install this plugin in DeepSeek Harness by running `dsh plugin install CC19990113/dsh-plugin-codegraph`, with the full source code available at https://github.com/CC19990113/dsh-plugin-codegraph.

About this plugin

When an AI agent prepares to edit code, it must first understand its underlying structure. However, traditional grep searches are easily distracted by comments and unrelated strings, while LSP relies on heavy language servers and specific cursor positions. To solve this, dsh-plugin-codegraph brings structural code intelligence directly into DeepSeek Harness, allowing the model to think about code topology just like a human developer.

The plugin equips the model with two core tools: codegraph and codegraph_index. Powered by a pre-built on-disk symbol graph, the agent can accurately answer complex questions in milliseconds, such as "who calls this," "what breaks if I change this," and "how do two symbols connect." It features a built-in Tree-sitter indexer that lazily loads grammars for over a dozen languages—including TypeScript, Python, Go, and Rust—and is fully interoperable with the database format of @colbymchenry/codegraph.

This plugin is ideal for developers using AI agents for large-scale refactoring, cross-file bug tracking, or anyone who wants their AI assistant to have a global "code map" before making changes. It moves beyond blind text matching, empowering the AI with true structural insight into your codebase.

Use Cases

  • Accurately assessing the impact scope of code modifications during cross-file refactoring
  • Tracing function call chains and dependencies in complex business logic
  • Enabling AI agents to quickly build a global code map before editing

Best For

  • Developers using AI agents for automated coding and refactoring
  • Engineering teams needing deep insights into large codebase structures
  • Plugin developers building local-first code intelligence tools