Repowise

What Repowise Does

Repowise is a codebase intelligence platform that creates a reusable, queryable model of your entire repository—combining source structure, git history, health signals, tests, and engineering decisions into one shared index. It gives AI coding agents precise context without resending the entire codebase, generates architecture documentation that stays current, flags risky changes before they merge, and surfaces code health issues validated against real defect history.

Who It's For & Problems It Solves

Repowise serves developers working with AI coding agents (Claude, Cursor, etc.), engineering teams collaborating on large codebases, team leads reviewing pull requests, and security teams triaging CVEs. The core problem: AI agents waste context windows rediscovering repository structure on every task, documentation falls out of sync with code, and teams lack visibility into which changes carry real risk. Repowise solves this by indexing once and reusing that intelligence across workflows.

Key Features & Capabilities

• AI Context via MCP: Ten Model Context Protocol tools that give agents queryable access to symbols, callers, decisions, and health signals—cutting agent output tokens by 31.6% in benchmarks
• Code Health Scoring: Defect-validated 1–10 scores per file, backed by real bug-fix history (0.74 ROC AUC across 21 repos), with drill-down to exact findings
• Change Risk Analysis: 0–10 risk score for any commit or PR, analyzing blast radius, test coverage gaps, and historical co-change patterns
• Auto-Generated Documentation: Architecture (C4 diagrams), module pages, and ADRs that rebuild when source changes, with freshness indicators and source links
• Git Intelligence: Hotspots, ownership, hidden coupling, bus factor, and reachability-aware CVE triage
• GitHub PR Bot: Automatically reviews diffs against the indexed codebase, surfacing changed contracts, missing co-changes, and merge-blocking issues

Typical Workflows

Developers connect Repowise via CLI or MCP to give their coding agent persistent codebase context. Teams install the GitHub App to get risk analysis on every pull request. Engineering leaders use the hosted dashboard to track AI-written code, health trends, and ownership gaps. Security teams prioritize CVEs based on whether vulnerable code is actually called.

Integrations & Deployment

Repowise integrates with GitHub (PR bot, repo browsing), MCP-compatible agents (Claude, Cursor), and command-line workflows. It runs three ways: open source (AGPL-3.0) for local/self-hosted use, hosted SaaS with free and paid tiers, or enterprise deployments (air-gapped, self-hosted, org-wide licensing).

What Sets It Apart

Repowise ranks #1 on ContextBench for retrieval efficiency and file coverage. Every score and prediction is traceable to repository evidence—no black-box LLM opinions. The same index powers agent context, documentation, risk analysis, and health scoring, eliminating the need for separate tools that each reprocess the codebase.