Multi-model refactoring analysis service. 3 AI models independently analyze code and produce structural improvement proposals as GitHub issues.
- Parse — Identify target repo, scope (file/module/full), and optional direction
- Analyze — 3 models with different perspectives examine the code in parallel
- Consolidate — Merge proposals, highlight consensus findings
- Issue — Create
[refactor]labeled issues on the target repo - Reflect — Record run, evolve prompts
Refactor ≠ Rewrite. Tests must pass before and after. We propose structural improvements, not behavior changes.
| Analyst | Model | Perspective |
|---|---|---|
| 🏗️ Arch | GPT-5.5 | Architecture, module boundaries, dependencies |
| 🔬 Lens | Claude Opus 4.7 | Code quality, readability, patterns |
| ⚡ Flux | Gemini 3.1 Pro | Efficiency, dead code, simplification |
This is a consultancy, not a contractor. We create issues — the target repo's agent decides what to implement.
Refactoring can have a direction: "make it more like Discord", "extract shared components", etc. This guides the analysts' focus beyond general cleanup.