Traycer works with the user to turn an objective into sequenced phases (essentially mini-prompts). It’s an interactive process where the agent may ask for clarification and develop high-level phases. The user can provide input to fine-tune and adjust the phases. Traycer works on each phase to create a detailed plan with file-by-file instructions covering low-level details about classes, variables, and call hierarchies. Each phase is then dispatched to the right agent (your choice) for code generation. After each handoff, Traycer runs verification: acceptance checks tied to your intent, tests, static analysis, and diff-level reasoning to auto-detect regressions and fix forward when possible.
