Autonomous Mode
For experienced teams that trust the workflow, autonomous mode auto-approves phase transitions unless the system detects critical issues. This lets the AI work through the full loop without waiting for human approval at each gate. Safety exceptions (security vulnerabilities, breaking API changes, data migration risks) always pause for human review regardless of mode.
How it works
Autonomous mode is configured per-project. When enabled, the framework evaluates each quality gate automatically. If the analysis, implementation checks, and audit all pass without safety exceptions, the workflow advances to the next phase. Safety exceptions are a configurable list of conditions that always require human review: security-related changes, database migrations, public API modifications, and dependency updates. You can customize the exception list and approval thresholds.
Configuration showing autonomous mode with safety exceptions
Screenshot coming soon
Why it matters
Manual approval at every gate is great for learning the workflow, but becomes a bottleneck for experienced teams. Autonomous mode preserves all the quality checks while removing the manual confirmation step. The AI still does the full analysis, writes specs, builds with TDD, and audits quality -- it just does not wait for you to say 'proceed' at each step. Safety exceptions ensure that high-risk changes always get human eyes, balancing speed with safety.
Workflow comparison showing manual vs autonomous mode timing
Screenshot coming soon
Related Features
Structured Development Loop
Analyze, Specify, Build & Test, Audit, Ship -- every feature goes through mandatory quality gates.
Quality Audit
A reviewer agent audits implementation against the original design for Architecture, Reliability, and Code quality.
Complexity Tiers
Automatically classifies work: Micro (single file), Light (2-phase), or Full (5-phase with all agents).