repowise 0.1.23

Codebase intelligence for developers and AI — generates and maintains a structured wiki for any codebase

repowise

Codebase intelligence for developers and AI.

repowise generates and maintains a structured, hierarchical wiki for any codebase. It keeps documentation accurate as code changes and exposes everything through an MCP server so AI coding assistants can query it in real time.

Features

• Automatic documentation — generates module, file, and symbol-level docs from source code
• Git intelligence — tracks churn hotspots, ownership, bus factor, and change patterns
• Dead code detection — finds confirmed unused exports, functions, and types
• Decision intelligence — captures why code is structured the way it is
• MCP server — 8 tools for AI assistants (Claude Code, Cursor, Windsurf, etc.)
• REST API + Web UI — browse the wiki, search, and explore architecture diagrams
• Codebase chat — ask questions about your codebase in natural language
• Multi-language — Python, TypeScript, JavaScript, Go, Rust, Java, C/C++, Kotlin, Ruby

Install

Claude Code Plugin (recommended)

The fastest way to get started — Claude handles installation, configuration, and indexing for you:

/plugin marketplace add repowise-dev/repowise-plugin
/plugin install repowise@repowise
/repowise:init

That's it. Claude walks you through choosing a mode, setting up your API key, and indexing your codebase. The plugin also auto-registers the MCP server and teaches Claude to use Repowise tools proactively.

pip

pip install repowise

All LLM providers (Claude, GPT, Gemini, LiteLLM) are included out of the box.

Quick Start

With Claude Code (interactive)

After installing the plugin, just run /repowise:init and Claude guides you through everything. Or ask naturally — "set up repowise for this repo" works too.

From the CLI

# Set your API key
export ANTHROPIC_API_KEY="sk-ant-..."   # or OPENAI_API_KEY, GEMINI_API_KEY

# Generate documentation for your codebase
cd /path/to/your-repo
repowise init

# Or skip LLM docs — just graph + git + dead code (free, <60 seconds)
repowise init --index-only

# Keep docs in sync after code changes
repowise update

# Start the MCP server for AI assistants
repowise mcp

# Browse the wiki in your browser
repowise serve

How It Works

1. Ingestion — parses every file using tree-sitter, extracts symbols, imports, and builds a dependency graph
2. Analysis — computes git signals (churn, ownership, recency), detects dead code, identifies architectural decisions
3. Generation — sends structured prompts to an LLM to produce wiki pages at every level of the hierarchy
4. Persistence — stores everything in SQLite (or PostgreSQL) with full-text and vector search
5. Serving — exposes the wiki through REST API, MCP server, and web UI

CLI Commands

Command	Description
repowise init	Generate full wiki documentation for a codebase
repowise update	Incrementally sync wiki after code changes
repowise watch	Auto-update wiki on file saves
repowise search	Full-text, semantic, or symbol search
repowise mcp	Start MCP server for AI editors
repowise serve	Start web UI server
repowise dead-code	Detect unused/unreachable code
repowise decision	Manage architectural decision records
repowise generate-claude-md	Generate CLAUDE.md for editor context
repowise export	Export pages to markdown/html/json
repowise reindex	Rebuild vector search index
repowise status	Show sync state and page counts
repowise doctor	Run health checks on wiki setup

MCP Tools for AI Editors

Once connected via repowise mcp, your AI editor gets 8 tools:

Tool	What it does
get_overview	Architecture summary, key modules, entry points, git health
get_context	Rich context for files/symbols — docs, ownership, decisions, freshness
get_risk	Modification risk — hotspot score, dependents, bus factor, trend
get_why	Why code is structured this way — decisions, git archaeology
search_codebase	Semantic search with git freshness boosting
get_dependency_path	How two modules connect through the dependency graph
get_dead_code	Tiered dead code report grouped by confidence
get_architecture_diagram	Mermaid diagram with optional churn heat map

Works with Claude Code, Cursor, Windsurf, Cline, and any MCP-compatible editor. For Claude Code, the plugin auto-registers the MCP server — no manual config needed.

Web UI

repowise includes a full web dashboard (Next.js + React + D3.js) with:

• Wiki browser — AI-generated docs with syntax highlighting, Mermaid diagrams, and git history sidebar
• Dependency graph — interactive force-directed visualization (handles 2000+ nodes)
• Codebase chat — ask questions about your code in natural language
• Search — full-text and semantic search with global command palette (Ctrl+K)
• Symbol index — searchable table of every function, class, and method
• Coverage dashboard — freshness breakdown with one-click regeneration
• Ownership view — contributor attribution and bus factor risk detection
• Hotspots — ranked high-churn files with commit history
• Dead code finder — unused code with confidence scores and bulk actions
• Decision tracker — architectural decisions with health monitoring

Starts automatically with repowise serve if Node.js 20+ is installed. No separate setup — the frontend is downloaded and cached on first run.

No Node.js? Use Docker instead:

docker build -t repowise https://github.com/RaghavChamadiya/repowise.git
docker run -p 7337:7337 -p 3000:3000 -v .repowise:/data repowise

See the User Guide for more options.

Requirements

• Python 3.11+
• Git (for repository analysis)
• An LLM API key (for documentation generation — not needed for analysis-only mode)
• Node.js 20+ or Docker (optional, for the web UI)

Documentation

• User Guide — complete CLI reference, web UI guide, workflows, and troubleshooting
• Architecture Guide — how the system is built and why each piece exists
• Core Library — ingestion, generation, persistence, providers
• CLI Package — all commands with every flag documented
• Server & MCP Tools — REST API endpoints, MCP tools, webhooks, scheduler
• Web Frontend — every page and component

Contributing

Contributions are welcome. Please read the Architecture Guide before submitting PRs.

# Development setup
git clone https://github.com/RaghavChamadiya/repowise.git
cd repowise
uv sync                                    # Python dependencies
npm install                                # Web frontend dependencies
pytest                                     # Run tests
ruff check packages/ tests/                # Lint

License

AGPL-3.0 — see LICENSE for details.