AI Coding & Dev2026-09-07•6 min read
Factory AI Droid Guide: Automating Codebase Refactoring with Frontier Models
How autonomous Droid agents navigate real repositories, generate unit tests, and execute complex feature specs without babysitting.
Unlike inline autocompletion extensions, Factory AI operates as an autonomous background engineer. You feed it a GitHub issue or specification, and its Droid agent clones your repo in an isolated sandbox, writes code, runs tests, and submits a verified PR.
Core Capabilities of Factory Droids
•**Multi-File Context:** Reads your complete dependency graph rather than isolated open tabs.
•**Automated Test Fixes:** Runs pytest, vitest, or cargo test inside a sandbox, diagnosing and repairing its own compiler errors.
•**Frontier Reasoning:** Powered by Claude 3.5 Sonnet and OpenAI reasoning models.
PerkShelf provides verified 1-year license codes for Factory AI Pro at $20 (MSRP $180/yr).