Introduction

In DeepSeek Harness (DSH), when performing code understanding, refactoring evaluation, or calling tools with agents, the common problem is not just “can a certain symbol be found”, but rather the position of the symbol within module dependencies, call hierarchies, and architectural boundaries. For example: which upstream references a specific function? Which files will deleting a module affect? Do two cross-language API call chains correspond? These questions require static analysis tools to provide relatively certain results.

Below is an introduction to DeepSeek Lens. It is a tool plugin for DeepSeek Harness named @trench-xinxin/dsh-tool-lens, maintained by trench-xinxin. It is positioned as a plugin for DeepSeek Harness focusing on Deep AST code graphs, symbol call hierarchies, and refactoring impact analysis.

Overview

DeepSeek Lens can be understood as a code structure analysis tool plugin within DSH. It provides a set of callable capabilities centered around the dependencies, calls, impact, and architecture relationships of a codebase, such as dependencies, call_graph, impact, circular, action: path, action: unused, action: lint, action: diff_impact, action: slice, and action: api_contracts.

It supports multiple language scenarios including Java, Python, Go, Rust, TypeScript, Vue 3, Svelte, etc. The license is MIT.

Core Capabilities

Based on verified data, DeepSeek Lens provides the following capabilities:

  1. Module and Dependency Topology Analysis (dependencies)
    Used to view static relationships between modules and dependencies.

  2. Scope Disambiguation Call Hierarchy Analysis (call_graph)
    Used to analyze symbol call hierarchies and reduce confusion caused by same-named symbols in different scopes.

  3. Layered Refactoring Impact Radius Analysis (impact)
    Used to evaluate the upstream and downstream scope that might be affected after modifying or removing a symbol.

  4. Circular Dependency Audit (circular)
    Used to check for the existence of circular dependencies in the code.

  5. Shortest Call Path Exploration (action: path)
    Used to view the call path between two symbols.

  6. Dead Code and Unused Symbol Audit (action: unused)
    Used to discover unreferenced symbols or isolated code.

  7. Architecture Layer Governance (action: lint)
    Used to check architectural boundaries or hierarchical dependency rules.

  8. Git Diff Impact Analysis and Test Recommendations (action: diff_impact)
    Used to perform analysis combined with code change impacts and provide test recommendations.

  9. Domain Architecture Subgraph Slicing (action: slice)
    Used to extract a subgraph around a specific domain or symbol from the code graph.

  10. Full-stack Cross-language API Contract Tracking (action: api_contracts)
    Used to track call relationships related to API contracts between frontend/backend or across languages.

Installation and Usage

The startup command provided in the documentation is:

npx @trench-xinxin/dsh-tool-lens

The README states that this command automatically downloads the runtime, handles port allocation, and launches the full Web UI with Lens. This means you first execute the command above, and then access the local address in your browser.

You can also run a prompt task directly in the terminal:

npx @trench-xinxin/dsh-tool-lens "Use lens tool to audit circular dependencies in this workspace"

After starting, the browser can access a local address, for example:

http://127.0.0.1:3080
http://127.0.0.1:3081

Additionally, the documentation mentions that peerDependenciesMeta will mark @deepseek-ai/cordis and @deepseek-ai/dsh-tools as optional.

Typical Usage

  1. Start the full Web UI
npx @trench-xinxin/dsh-tool-lens

This step launches the local Web UI with Lens. After starting, access the page according to the terminal prompts or the local address.

  1. Execute a task directly in the terminal
npx @trench-xinxin/dsh-tool-lens "Use lens tool to audit circular dependencies in this workspace"

This step is suitable for having Lens perform a circular dependency audit on the current workspace directly in the terminal.

  1. Access the local Web UI

After starting, access a local address similar to the following in your browser:

http://127.0.0.1:3080

Or:

http://127.0.0.1:3081

The actual address depends on the terminal prompts at startup.

Applicable Scenarios and Notes

Suitable for the following scenarios:

  1. Adding code structure analysis capabilities to a DSH workflow.
  2. Evaluating the refactoring impact radius before refactoring.
  3. Auditing circular dependencies, unused symbols, or architectural boundaries.
  4. Performing code graph analysis in multi-language projects such as Java, Python, Go, Rust, TypeScript, Vue 3, Svelte, etc.

Notes to be aware of before use:

  1. After the plugin is enabled in DeepSeek Harness, it will read code and execute analysis with the permissions of the current dsh process.
  2. Before installing or enabling, you should check the source code, dependencies, license, and runtime permissions.
  3. This plugin is declared as MIT License © 2026 Trench / DeepSeek Lens Contributors.
  4. The command provided in the documentation is the npx start/run command and is not separately labeled as an install command.
  5. This plugin belongs to the DeepSeek Harness plugin ecosystem; the community directory is an independent site and should not be understood as an official app store.

Conclusion

The value of DeepSeek Lens is to integrate capabilities such as code dependencies, call hierarchies, refactoring impact, circular dependencies, and API contract analysis into DeepSeek Harness. For developers who need to perform code understanding and refactoring evaluation within agent workflows, it provides a tool entry point that can be started directly via npx.

Directory page:
https://www.skillhub.cn/plugins/trench-xinxin/dsh-tool-lens

GitHub:
https://github.com/trench-xinxin/dsh-tool-lens