Author: Michael Tittmar
Rule Intelligence MCP is a flexible open-source CLI & library to analyze, validate, and manage any kind of rulebase – for any language, team, or workflow. It supports multiple formats, custom checks (plugins), AI integration, and more.
- Rulebase Analysis: Find missing categories, duplicate rules, empty content, and other issues in your rules. Example: Detect if a coding guideline is missing a category or if two rules have the same title.
- Schema Validation:
Check your rules against any JSON schema to ensure consistency and completeness. Example: Enforce that every rule has an
id,title, andcategory. - Flexible Filtering: Filter rules by status (e.g. active, deprecated), tags, or custom fields. Example: List only active security rules for a release checklist.
- Multiple Output Formats: Export results as JSON (for automation), Markdown (for docs), CSV (for Excel), or HTML (for reports). Example: Generate a Markdown summary for your team wiki.
- Interactive Editing: Edit rules directly from the CLI, e.g. update a title or category without opening a text editor.
- Memory Bank (Change History): All edits and suggestions are logged with timestamps. Example: Track who changed what and when.
- AI Integration (LLM): Get suggestions for new or improved rules from AI models like OpenAI GPT. Example: "Suggest 3 new security rules for my project."
- Live Watch Mode: Automatically re-analyze your rulebase when files change. Example: Get instant feedback while editing.
- Plugin System: Write your own checks/analysis in JavaScript. Example: Plugin that finds conflicting rules.
- Flexible Input: Supports JSON, YAML, TOML, and Markdown (frontmatter). Example: Use the tool for rules in code, docs, or config files.
- Tested & CI-ready: Includes unit and integration tests, ideal for CI pipelines to enforce rule quality before merging code.
- Node.js ≥ 14
- npm or yarn
git clone https://github.com/micha-gh/rule-intelligence-mcp.git
cd rule-intelligence-mcp
npm installrule-intelligence-mcp <command> [options]| Option | Description | Example |
|---|---|---|
-r, --rulebase |
Path to rulebase file (JSON/YAML/TOML/MD) | -r rules.json |
-s, --schema |
Path to JSON schema file | -s my-schema.json |
--format-in |
Input format: json, yaml, toml, md (usually auto-detected) | --format-in yaml |
-o, --output |
Output format: json, md, csv, html | --output csv |
--status |
Filter rules by status | --status active |
--tag |
Filter by tags (comma-separated) | --tag security,style |
--plugin |
Path to JS plugin for custom checks | --plugin ./plugins/my-check.js |
--llm-provider |
LLM provider (openai, azure, local, ...) | --llm-provider openai |
--llm-api-url |
LLM API base URL (for custom providers) | --llm-api-url https://api.example.com |
--llm-model |
LLM model name | --llm-model gpt-4 |
--memory |
Path to memory bank file | --memory .data/mem.json |
# Analyze a rulebase (format usually auto-detected)
rule-intelligence-mcp analyze -r myrules.json --output md
rule-intelligence-mcp analyze -r myrules.yaml --format-in yaml --output csv
rule-intelligence-mcp analyze -r myrules.json --plugin ./plugins/example-plugin.js --output json
# Edit a rule
rule-intelligence-mcp edit rule-2 --title "Use parameterized queries"
# Show memory bank
rule-intelligence-mcp memory:list --memory .data/mem.json
# AI suggestions (optional, install openai first)
npm install openai
rule-intelligence-mcp suggest -r myrules.json --llm-provider openai --llm-model gpt-3.5-turbo
# Watch mode (optional, install chokidar first)
npm install chokidar
rule-intelligence-mcp watch -r myrules.json- Use any JSON schema for validation (
--schema my-schema.json) - Rulebases can be JSON, YAML, TOML, or Markdown (with frontmatter)
- Format is usually auto-detected, but can be set with
--format-in - YAML example:
rule-intelligence-mcp analyze -r rules.yaml --format-in yaml
- Markdown example:
rule-intelligence-mcp analyze -r rules.md --format-in md
--output json(default, for automation)--output md(Markdown, for docs)--output csv(Excel/tables)--output html(reports)
- Write your own checks as a JS file:
--plugin ./my-check.js - The plugin receives the rulebase and the current analysis result as input and returns an object to merge into the output
- See
plugins/example-plugin.jsas a template - Example: Count deprecated rules and add the result
- Use any LLM provider: OpenAI, Azure, local, etc.
- Flags:
--llm-provider,--llm-api-url,--llm-model - Example:
rule-intelligence-mcp suggest --llm-provider openai --llm-model gpt-4 - Install required packages first (see docs)
- Set API keys as environment variables (e.g.
OPENAI_API_KEY)
- Stores all
editandsuggestinteractions with timestamps - Location and backend are configurable (
--memory memory.json,--memory .data/mem.json) - Manage with
memory:listandmemory:clear
- Run all tests:
npm test - Coverage:
npm test -- --coverage - GitHub Actions workflow in
.github/workflows/ci.yml
- The
examples/folder contains various rulebases (minimal, complex, invalid, markdown) - The
plugins/folder contains example plugins - The
test/folder contains test scripts for positive and negative cases
Run automated tests:
bash test/examples.test.sh
bash test/examples.error.test.shTest matrix and notes can be found further down in the README.
You can use Rule Intelligence MCP as a Custom Tool in Cursor AI. This allows you to analyze and validate rulebases directly from your IDE.
Step-by-step guide: docs/cursor-integration.md
Want to help? See CONTRIBUTING.md for how to open issues, submit PRs, and coding guidelines.
Please read and follow our Code of Conduct.
See CHANGELOG.md for release notes and version history.
MIT © Michael Tittmar