HoloGram DSH plugin

HoloGram compiles a codebase into a unified IR dependency graph using tree-sitter static analysis. It exposes 34 graph query tools through the MCP protocol, enabling AI agents to reason about dependencies deterministically rather than by guessing. The DSH plugin bundles the engine with a 3D star map visualization that renders the project dependency graph directly in the DSH web interface.

developer-tools vision

How to install the HoloGram DSH plugin

dsh plugin --profile web add @a834063245/hologram-dsh

Copying does not run this command. Review the repository and version before installing the HoloGram DSH plugin.

HoloGram DSH plugin data source

HoloGram DSH plugin snapshot date: Aug 16, 2026

discovered

What the HoloGram DSH plugin can do

34 MCP tools for dependency queries, risk analysis, architecture diagnostics, execution flow, and data flow (heuristic).
3D star map visualization (Three.js WebGL) accessible via DSH web sidebar.
Supports 27 languages via tree-sitter grammars, with 18 languages having hand-tuned structural extraction.
Incremental updates via watcher; single binary engine with zero configuration.
Preflight check, cycle detection, impact trace, and fragile module identification.

Where the HoloGram DSH plugin fits

Analyze the impact of changing a function or module before committing.
Detect circular dependencies and fragile modules in large codebases.
Understand code architecture and data flow paths for onboarding or refactoring.
Integrate deterministic dependency reasoning into AI coding assistants.

Who the HoloGram DSH plugin is for

AI agent developers who need reliable codebase understanding.
Software architects and engineers maintaining large, complex codebases.
Teams using DSH (DeepSeek Harness) and seeking advanced static analysis capabilities.

HoloGram DSH plugin limitations

Only supports Windows and Linux (no macOS support as of the current release).
Dynamic code patterns (eval, dynamic imports) are marked as blind spots and not fully resolved.
Requires an AI coding tool or DSH environment to utilize the MCP tools; the engine alone is a CLI/MCP binary.
Large project analysis can take significant time (e.g., Linux kernel ~1770s) and memory (RSS ~646MB).

HoloGram DSH plugin: from the repository README

Quoted from the 834063245-creator/HoloGram README, the upstream source of the HoloGram DSH plugin. Copyright remains with the original authors.

HoloGram 把代码库编译成一张统一 IR 依赖图（节点=符号/函数/类/模块，边=调用/继承/读写/时序），并通过 MCP 协议向 AI Agent 暴露 34 个图查询工具（注册表 35 个 schema，`symbol_history` 为 legacy 不默认暴露）。 **核心主张：依赖推理应当是确定性的，而不是猜的。** LLM 分析"改 A 会炸什么"时，靠逐文件读源码推测依赖——弱模型会漏，大项目会翻不动。HoloGram 用 tree-sitter 静态分析预先算好整张依赖图：Agent 一次工具调用拿到结构化事实（影响面、循环、脆弱模块、数据流路径），而不是源文件文本。单点查询省 ~70% token，全局分析省 90%+；省 token 是次要的，**可靠性**是主要的。引擎是单文件二进制，本地运行、零配置、代码不出机器。 ## 快速开始 ### MCP 模式（推荐，1 分钟）引擎随 [Releases](https://github.com/834063245-creator/HoloGram/releases) 发布（Windows / Linux），安装脚本一键完成。也可以把下面这段话直接发给你的 AI 编程工具，让它自己装： ``` 请帮我安装 HoloGram MCP 服务。步骤： 1. 从 https://github.com/834063245-creator/HoloGram/releases 下载： - Windows: hologram-engine-windows-x64.zip - Linux: hologram-engine-linux-x64.tar.gz 2. 解压后运行安装脚本： - Windows: 双击 install.cmd - Linux: ./install.sh --user 3. 在当前 AI 编程工具的 MCP 配置中注册： - command: hologram-engine - args: serve 4. 重启 AI 编程工具，调 engine_status 验证 ``` <details> <summary>手动配置</summary> **Claude Code** — `~/.claude/mcp.json`： ```json { "mcpServers": { "hologram": { "command": "hologram-engine", "args": ["serve"] } } } ``` **Cursor** — Settings → MCP → Add new MCP server：command `hologram-engine`，args

Read the full READMERepository license: MIT

HoloGram DSH plugin questions

How do I install the HoloGram DSH plugin?

Run `dsh plugin --profile web add @a834063245/hologram-dsh` and then restart DSH web. The plugin provides 34 MCP tools and a 3D star map sidebar entry. Ensure you have a DSH environment (version supporting plugin bundles).

What languages does HoloGram support?

It supports 27 languages via tree-sitter grammars. 18 languages (Python, TypeScript, Rust, Go, Java, C/C++, C#, Ruby, PHP, Swift, Dart, Scala, Zig, Elixir, Lua, Bash, R) have hand-tuned structural extraction. Others (OCaml, Haskell, Nix, etc.) fall back to generic grammar parsing. Kotlin, Markdown, TOML are dynamically loaded.

Does HoloGram work with macOS?

No, the current release only provides prebuilt binaries for Windows and Linux. macOS is not mentioned as supported. You may need to build from source or wait for future releases.

Can I use HoloGram without DSH?

Yes, HoloGram also runs as a standalone CLI/MCP binary. Download from Releases, run `hologram-engine serve`, and configure it in your AI coding tool (e.g., Claude Code, Cursor). The DSH plugin is an optional bundle for DSH users.

How does HoloGram handle dynamic code like eval or dynamic imports?

HoloGram marks these patterns as blind spots. It does not pretend to know runtime-only facts. Dynamic imports are marked as dynamic sites, and eval is marked as unreachable. This ensures the analysis is honest about its limitations.