We have 2 major workflows.
1. Plan mode - you simply start with a task query and then Traycer explores the codebase, generating a file-level detailed plan with proper descriptions, references, mermaid diagrams, reasoning, etc. You can then iterate on the plan or chat with it to ask anything. Once your plan is final, you can execute it using any of your favorite coding agents like Cursor, Claude Code, Github Copilot, and others. After execution, these agents will follow the plan properly. You can then return to Traycer and click verify changes; Traycer will check if all modifications made by agents adhere to the plan. If any items are missing, Traycer suggests changes directly on the UI in the form of comments. You can use these comments to hand off the fixes to your agent.
2. Phase mode - this is for a step-by-step approach. You add a query, and Traycer understands your intent and asks questions if needed. It then creates multiple phases. You generate a plan for each phase, similar to Plan Mode, where you get everything. After verifying a phase, you move on to the next, keeping the context of previous phases.
Here's a quick video by a developer: