Repowise

Repowise is an open-source codebase-intelligence layer for software teams and coding agents. It indexes a repository's code, call graph, git history, tests, docs, design decisions, dead code signals, and code health metrics once, then answers questions that would otherwise force an agent to rediscover the project by reading huge chunks of source. The public README frames the product around a simple pain point: agents burn tokens and make risky edits when they do not know what calls a function, which tests cover a file, why a design choice was made, or what technical debt should be fixed first. The tool is built for local, evidence-backed repository analysis. Repowise reports call graph quality, code health scores, dead code detection, git analytics, auto-generated docs, architectural decisions, and risk-focused views. Its repository says the graph, risk, health, tests, and dead-code analysis make zero LLM calls, need no API key, and can run self-hosted. That matters for teams that want agent context without sending every file through a model or paying for repeated summaries. The project also advertises a live demo and a PyPI package, with MCP support so AI clients can query the index as a structured tool surface. For daily developer work, Repowise is most useful when a repository is too large for a single prompt. A developer or agent can ask what calls a function, what may break if a file changes, where tests exist, which files look risky, which code is probably dead, and why a part of the system was built that way. That makes it useful for refactoring, onboarding, pull-request review, architecture cleanup, and technical-debt planning. The README publishes benchmark-style claims, including less agent output, fewer tool calls, and more surfaced defects in selected comparisons. Those claims should be read as project-provided measurements, not neutral third-party results, but they explain the intended value. Repowise fits teams using Claude Code, Cursor, Codex, or other coding agents that need a better memory layer over source code. It is not a replacement for running tests, code review, or human architecture judgment. It is also not a hosted closed SaaS by default; the repository is AGPL-3.0 with commercial licensing noted, and the public path is free and self-hosted. Teams adopting it should review the license, decide where indexes live, and verify language support for their stack before depending on it in CI. The strongest OpenTools angle is cost control for agentic software work. Instead of asking a model to read the same files repeatedly, Repowise builds reusable repository intelligence and exposes it through docs, badges, dashboards, and MCP tools. That can reduce wasted model output and give humans a shared view of what is risky before they merge.