AI Agent Hub
Back to plugins
🖥️

dsh-codegraph-plugin

Client Updated 2026.08.31

Run the following command in DeepSeek Harness:

dsh plugin install errrepe/dsh-codegraph-plugin

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

Run dsh plugin install errrepe/dsh-codegraph-plugin in your terminal; the source is at https://github.com/errrepe/dsh-codegraph-plugin — after installation, restart the DSH process to activate the plugin.

About this plugin

When an agent in a large codebase relies on grep to hunt for text, it typically gets back a wall of unrelated snippets with no structural context. dsh-codegraph-plugin wraps the codegraph CLI into eight code-intelligence tools so the model can query symbol dependencies, call graphs, and impact scope directly, replacing blind text search.

The plugin always operates on the workspace of the calling session and keeps a local index in sync. The eight tools cover symbol lookup, caller/callee queries, impact analysis, code browsing, and index status checks. After editing files a single sync call brings subsequent queries up to date. All dynamic arguments are POSIX-safe quoted, and the SQLite WAL index works inside DSH sandbox policies.

It also plugs into DSH Settings > Plugins: one toggle enables or disables all eight tools, and a second toggle controls whether a guidance section is injected into the system prompt, instructing the model to prefer graph queries over grep. Both flags persist independently under the codegraph namespace in ~/.dsh/settings.yaml with no residual output when turned off.

Built for developers who use DSH on real codebases and want their agent to reason about code structure rather than raw text. The host machine must already have the codegraph CLI installed and available on PATH.

Use Cases

  • Quickly look up which modules call a given function in a large project
  • Assess the impact scope of a core interface before modifying it
  • Have the agent reason over dependency structure instead of line-by-line text search

Best For

  • Developers who work with real codebases inside DSH
  • Teams that need their agent to have structured code comprehension
  • Prompt engineering practitioners who prefer graph queries over blind grep