Repowise
Repowise is an open-source, self-hostable codebase intelligence platform that provides AI coding agents and development teams with a verified index of a repository's structure, history, and risk. By analyzing the codebase and its Git history, it builds a queryable model that agents can access via the Model Context Protocol (MCP), drastically reducing the need for agents to guess at context and minimizing human review bottlenecks. The platform offers a comprehensive suite of features including a defect-validated code health score per file, agent provenance tracking to monitor AI-written code, change risk assessment for commits and pull requests, reachability-aware CVE triage, an auto-generated wiki, Git intelligence such as hotspots and bus factor, architecture diagrams in C4 model, and mining of architectural decisions from multiple sources. These tools collectively address the three main leaks in AI velocity: agents guessing, humans re-verifying, and teams forgetting institutional knowledge. In benchmarks, Repowise demonstrated a 96% reduction in tokens required to load context for AI tasks, while maintaining answer quality. It also caught 2.3 times more defects per review compared to a leading commercial tool under the same budget. The code health score is validated against real defects across projects, achieving a cross-project ROC AUC of 0.74, and up to 0.90 per repository. With support for incremental updates in under 30 seconds, it integrates seamlessly into CI/CD pipelines and developer workflows. Designed for both individual developers and large enterprises, Repowise can be installed via pip and self-hosted, ensuring data privacy and control. A hosted version is available for convenience, with a free tier for public repositories and Pro plans for private repos. The open-source core and permissiveness encourage community contributions and customization.
- Category
- AI Agents
- Pricing
- Free
Problem solved
AI coding agents often lack deep context about a codebase, leading to incorrect changes, excessive review cycles, and loss of institutional knowledge. Repowise solves this by indexing a repository's structure, history, and risk into a queryable model, providing agents with the necessary context to make informed decisions, and giving teams insights into code health, change impact, and security vulnerabilities.
Use cases
- Providing AI coding agents (like Claude Code or Cursor) with pre-indexed context via MCP to reduce token usage and errors
- Assessing code health with a defect-validated score to prioritize refactoring
- Tracking AI-written code provenance and its impact on code quality
- Evaluating change risk for any commit or pull request to prevent defects
- Triaging CVEs by determining if vulnerable code is actually reachable in your dependency graph
- Auto-generating and maintaining a wiki of your codebase that evolves with changes
- Identifying knowledge hotspots, bus factor, and hidden coupling to manage team risks
- Mining architectural decisions (ADRs) from code, commits, and discussions for better documentation
Audience
- Developers
- Enterprise
Pros
- Open source and self-hostable ensuring data privacy and control
- Reduces AI context token usage by up to 96% while maintaining answer quality
- Catches 2.3x more defects per review compared to leading commercial tools
- Validated code health score with cross-project ROC AUC of 0.74
- Comprehensive feature set covering code health, security, architecture, and documentation
Cons
- Requires self-hosting or a Pro plan for private repositories
- Initial indexing of large repositories may take time
- Limited to Git version control systems
- May require integration setup with MCP clients for full AI context features
- Enterprise features under commercial license may incur costs
Official website: https://www.repowise.dev/
Canonical LambdaBase page: https://www.lambdabase.com/tools/repowise