codegraph-mcp vs repowise
No-train, on-prem code knowledge graph served to AI agents over MCP — symbols, call edges, cross-language links and blast-radius queries, with a hash-chained audit log of every read. — versus — Codebase intelligence for AI and humans: deterministic code-health scores calibrated on real defects, graph-aware refactoring plans agents can execute, auto-docs and git analytics over 9 MCP tools.
Both serve a code knowledge graph to agents over MCP; codegraph-mcp goes deep on cross-language symbol/call/blast-radius queries, repowise trades some of that depth for defect-risk scores and refactoring plans.
| codegraph-mcp | repowise | |
|---|---|---|
| Stars | 5 | 3.7k |
| Forks | 0 | 438 |
| Language | Python | Python |
| License | NOASSERTION | AGPL-3.0 |
| Last activity | 12 days ago | yesterday |
| Topics | coding, rag | coding |
| Curated connections | 4 | 3 |
codegraph-mcp — the curator's take
Niche but real: if agents must understand code that cannot leave your building AND compliance asks 'what exactly did the agent read', the tamper-evident audit chain is the only game in town; cross-language call edges (TS fetch → Go/Python handler) catch what one-file context misses. NOT for most teams yet: 5 stars, single vendor, and — deal-breaker until fixed — no clear open-source license (NOASSERTION on GitHub). Treat it as an evaluation candidate, not a dependency.
repowise — the curator's take
The interesting bet: defect-risk scoring with NO LLM — 25 deterministic markers calibrated against a real defect corpus (published ROC AUC 0.74), indexed in seconds, then the same dependency graph generates concrete refactoring plans (split the god class, break the cycle) your coding agent executes. Health→locate→fix as one loop is what linters and dashboards don't do. NOT a pure-open play: AGPL-3.0 with a hosted-teams funnel, and benchmark claims are self-published — reproduce them on your repo before quoting them; overlaps only partially with symbol-level code-graph MCPs.