MCP Server

Active

MIT

Hex Research MCP Server

io.github.levnikolaevich/hex-research-mcp

Engineering skills marketplace for Claude Code and Codex with audits, reviews, optimization, testing, and architecture workflows.

Repository Registry

What is the Hex Research MCP server?

The Hex Research MCP server is a research graph for managing hypotheses, goals, runs, source quality, audits, and generated maps. It provides structured data management capabilities for tracking research workflows and evidence quality across engineering projects.

This server exposes a research graph infrastructure designed to support hypothesis-driven engineering work. It enables tracking of research goals, experimental runs, source quality metrics, audit results, and generated documentation maps—useful for teams building reproducible engineering workflows and maintaining evidence-backed decision records.

How to install Hex Research

Copy-paste configuration for popular MCP clients.

transport: stdio

Config generated by PluginBench — verify against the source before use.

~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "hex-research-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@levnikolaevich/hex-research-mcp"
      ]
    }
  }
}

Tools & capabilities

Tools this server exposes to the agent.

Research graph management — Store and query hypotheses, goals, runs, and research artifacts
Source quality tracking — Assess and track the quality and trustworthiness of evidence sources
Audit recording — Record and manage audit results and findings
Map generation — Generate and manage research or architecture maps from graph data

Use cases

Track hypothesis validation across multiple experimental runs and record evidence quality
Audit codebase, test suites, documentation, and architecture decisions with structured findings
Generate architecture diagrams and documentation maps from repository evidence
Plan and execute reversible changes with reproducible experiment contracts
Review implementation plans and completed code through independent risk-selected perspectives

Hex Research MCP server FAQ

What is the Hex Research MCP server?

It is a research graph MCP that manages hypotheses, goals, experimental runs, source quality assessments, audits, and generated maps to support evidence-driven engineering workflows.

Is it free?

Yes, the repository is licensed under MIT and available on GitHub at github.com/levnikolaevich/claude-code-skills.

How do I install it?

Install via npm: npm install @levnikolaevich/hex-research-mcp. For Claude Code, add the marketplace with /plugin marketplace add levnikolaevich/claude-code-skills and install individual suites. For Codex, use codex plugin marketplace add and codex plugin add commands.

What skills does it include?

Seven suites: Review Suite (plan and delivery review), Codebase Audit Suite (documentation, code, tests, architecture, persistence), Optimization Suite (performance, dependencies, modernization, benchmarks, surgical changes), Testing Suite (strategy and acceptance tests), Product Discovery, Maintainer Suite, and Architecture Suite (design, documentation, proposals, decisions, diagrams, migration planning).

Does it require authentication?

No authentication is mentioned in the documentation. It operates as a local plugin marketplace for Claude Code and Codex.

Can I use individual skills?

Yes, you can install only the suites you need. Each skill is invoked by its namespaced name, for example /review-suite:ln-12-delivery-reviewer.

README (reference)

Source of truth, from the repository.

Claude-Codex Skills

A compact marketplace of standalone engineering skills for Claude Code and Codex.

Supports the portable Agent Plugins v1 standard while retaining native Claude Code and Codex distribution compatibility.

Why this repository is intentionally small: Earlier coding models needed a large orchestration and evaluation harness to follow complex workflows reliably. Modern Claude and Codex models work better with concise procedural guidance, so that machinery has been removed. These skills retain only the domain knowledge, decision gates, tool guidance, and evidence checklists worth bringing into context.

Browse the skills catalog or install only the plugins you need below.

The repository intentionally contains only the skills, minimal plugin manifests, two host-specific marketplace catalogs, documentation, and a static catalog site. It has no MCP servers, orchestration hierarchy, distributed shared resources, generated skill copies, or evaluation harness.

Plugins

Review Suite

Index	Skill	Purpose
11	Plan Reviewer	Validate an implementation plan against repository evidence before execution.
12	Delivery Reviewer	Review completed code through independent risk-selected perspectives.

Codebase Audit Suite

Index	Skill	Purpose
21	Documentation Auditor	Audit documentation trust, coverage, structure, and factual accuracy.
22	Codebase Auditor	Audit cross-cutting code health, security, delivery, and maintainability.
23	Test Suite Auditor	Audit test value, coverage, isolation, and oracle strength.
24	Architecture Auditor	Audit system boundaries, ownership, contracts, and dependency topology.
25	Persistence Auditor	Audit queries, transactions, consistency, and resource lifecycle.

Optimization Suite

Index	Skill	Purpose
31	Performance Optimizer	Profile, experiment, and keep only measured improvements.
32	Dependency Upgrader	Upgrade dependencies in reversible, verified batches.
33	Code Modernizer	Replace or simplify bounded capabilities when net value is proven.
34	Benchmark Comparator	Compare alternatives with a frozen, reproducible experiment contract.
35	Surgical Change Implementer	Implement a bounded change through the smallest complete root-cause solution.

Testing Suite

Index	Skill	Purpose
41	Test Strategy Planner	Design a risk-based, decision-complete test strategy without changing code.
42	Acceptance Test Builder	Create reproducible acceptance tests through observable product boundaries.

Product Discovery Suite

Index	Skill	Purpose
51	Opportunity Evaluator	Compare product opportunities using current evidence and a low-cost validation path.

Maintainer Suite

Optional maintainer toolkit: skill 61 reviews skill repositories, skill 62 publishes any Git repository with equivalent remote evidence, and skills 63–64 publish GitHub releases and Discussions. Users who only consume the engineering skills do not need to install it.

Index	Skill	Purpose
61	Skill Reviewer	Review standalone skills and configured distribution surfaces before publication.
62	Repository Publisher	Validate, commit, push, and remotely verify approved repository changes.
63	Release Publisher	Prepare and publish an approved tagged GitHub release.
64	Community Announcer	Draft and publish fact-checked GitHub Discussions project announcements.

Architecture Suite

Creates durable architecture artifacts without coupling the skills to one another. Each workflow operates independently and communicates only through optional repository documents.

Index	Skill	Purpose
71	System Design Baseline Builder	Establish measurable architecture drivers and constraints in one project baseline.
72	Current Architecture Documenter	Document implemented architecture from repository evidence.
73	System Design Proposal Builder	Turn requirements and constraints into a decision-complete target design.
74	Architecture Decision Recorder	Record one significant decision with alternatives and consequences.
75	Architecture Diagram Builder	Create evidence-backed current or target architecture views.
76	Architecture Migration Planner	Plan a reversible current-to-target transition with compatibility and rollback.

Install in Claude Code

Add the marketplace and install only the suites you need:

/plugin marketplace add levnikolaevich/claude-code-skills
/plugin install review-suite@levnikolaevich-skills-marketplace
/plugin install codebase-audit-suite@levnikolaevich-skills-marketplace
/plugin install optimization-suite@levnikolaevich-skills-marketplace
/plugin install testing-suite@levnikolaevich-skills-marketplace
/plugin install product-discovery-suite@levnikolaevich-skills-marketplace
/plugin install maintainer-suite@levnikolaevich-skills-marketplace
/plugin install architecture-suite@levnikolaevich-skills-marketplace
/reload-plugins

Invoke a skill by its namespaced name, for example /review-suite:ln-12-delivery-reviewer.

For local development, load one plugin directly:

claude --plugin-dir ./plugins/review-suite

Install in Codex

codex plugin marketplace add levnikolaevich/claude-code-skills
codex plugin add review-suite@levnikolaevich-skills-marketplace
codex plugin add codebase-audit-suite@levnikolaevich-skills-marketplace
codex plugin add optimization-suite@levnikolaevich-skills-marketplace
codex plugin add testing-suite@levnikolaevich-skills-marketplace
codex plugin add product-discovery-suite@levnikolaevich-skills-marketplace
codex plugin add maintainer-suite@levnikolaevich-skills-marketplace
codex plugin add architecture-suite@levnikolaevich-skills-marketplace

Repository layout

.
├── .agents/plugins/marketplace.json       # Codex catalog
├── .claude-plugin/marketplace.json        # Claude Code catalog
└── plugins/
    ├── review-suite/
    ├── codebase-audit-suite/
    ├── optimization-suite/
    ├── testing-suite/
    ├── product-discovery-suite/
    ├── maintainer-suite/
    └── architecture-suite/

Each plugin contains a portable Agent Plugins v1 plugin.json, the current .codex-plugin/plugin.json OpenAI host adapter, and one shared skills/<skill>/SKILL.md tree.

This is the smallest practical shared layout for distributed plugins:

Both hosts use skills/<name>/SKILL.md, so each skill has one canonical copy.
Agent Plugins clients discover the portable package through root plugin.json; its minimal manifest owns only the schema target and stable name.
Current ChatGPT and Codex packaging still requires .codex-plugin/plugin.json, which remains the single owner of mutable version, description, publisher, and interface metadata.
Claude Code scans each marketplace source's standard skills/ directory, so a duplicate Claude-specific plugin manifest is unnecessary.
agents/openai.yaml, references, scripts, assets, hooks, agents, and MCP configuration are optional and omitted until a concrete need appears.

The structure follows the Agent Plugins v1 specification, current OpenAI plugin guide, Agent Skills specification, Claude Code skill guide, and Claude Code plugin reference.

Indexing

The first digit identifies the plugin and the second identifies the skill within it: 1x review, 2x audit, 3x optimization, 4x testing, 5x product discovery, 6x repository maintenance, and 7x architecture artifact creation. See the canonical allocation and overflow rules in AGENTS.md.

License

MIT

Related MCP servers

Hex SSH

Active

Token-efficient SSH MCP with hash-verified files and persistent tmux sessions for remote code execution.

538

HTML

MIT

View repository →

Hex Graph

Active

Portable engineering skills marketplace for Claude Code and Codex with deterministic code analysis and architecture workflows.

538

HTML

MIT

View repository →

Hex Line

Active

Portable engineering skills marketplace for Claude Code and Codex with modular code review, audit, optimization, and architecture workflows.

538

HTML

MIT

View repository →

io.github.lewo-media/safe-mind

Maintained

Psychological risk assessments (GBU Psyche) via Safe Mind: surveys, departments, analytics.

TypeScript

MIT

View repository →

Probe V0

Active

Experimental MCP server for current empirical verification of explicit public HTTPS endpoint claims.

MIT

View repository →

io.github.lexerAnn/kindlesms

Send SMS, manage approved sender IDs, and check credits via KindleSMS.