DevFix AI — Investigate. Fix. Verify

DevFix AI — Investigate. Fix. Verify. Software debugging often requires developers to spend significant time understanding an unfamiliar repository, tracing an issue across multiple files, identifying the root cause, planning a safe fix, applying changes, and repeatedly testing the result. This can lead to unnecessary investigation, rework, and verification effort. DevFix AI is an AI-assisted developer workflow designed to turn this process into a structured, verifiable pipeline. A developer provides a software repository and a problem. DevFix AI then investigates the repository, analyzes the issue, identifies the root cause, creates a fix plan, waits for developer approval, applies the approved fix, and verifies the result. The workflow is organized around six specialized stages: Repository Explorer, Root Cause Investigator, Fix Planner, Developer Approval, Fixer Agent, and Verification Agent. An Orchestrator coordinates these stages and passes the results from one step to the next. A key part of the solution is the human approval boundary. DevFix AI does not directly apply proposed changes without developer approval. This keeps the developer involved in the decision before the fixing stage begins. For our demonstration, DevFix AI investigates an intermittent 401 Unauthorized error after token refresh in a sample API gateway repository. The system traces the request and token lifecycle, identifies a race condition, produces a structured fix plan, applies the approved solution, and verifies the result through build, unit-test, regression-test, and lint checks. The final report presents the outcome as FIX VERIFIED. The project uses a modern web architecture with React, TypeScript, Vite, Tailwind CSS, Node.js/API abstraction, specialized agents, and local state management. Demo/Mock execution provides a consistent end-to-end demonstration of the workflow. IBM Bob 2.0 was used during development as an AI development partner for project setup, agent architecture, workflow implementation, API abstraction, debugging, build/type-checking activities, and documentation. DevFix AI demonstrates how a complex debugging workflow can be divided into specialized AI-assisted stages while keeping the developer in control. Investigate. Fix. Verify.