An LLM editing your code is a guess. act101 makes it a transform: your agent refactors or ports the code, then act gate returns a deterministic MERGE / REVIEW / BLOCK verdict with a signed receipt and one-command rollback. 163 grammars, 183 AST refactor operations, 41 codebase analyzers, exposed via MCP (Claude Code, Cursor, Codex, OpenCode) and the act CLI — one native binary. Agent refactor Python. Agent port C to Rust. Agent port Ruby to Elixir.

I was cooped up in an RV somewhere in the Mohave Desert when a week's of work came to a screeching halt with the realization that my agents had left the user dimension off of half of the API surface because we weren't consuming it at that time. Eight hours and millions of tokens later, it was complete. And it was correct. But it wasn't right. After we adopted refactoring processes and tools, codebases became long-term, maintainable investments. No more 5-year rewrites. Agents need the same tooling that engineers and architects use. Agents perform more like engineers and architects when they have those tools. So I built act101: the definitive codebase quality and refactoring tool for coding agents.