One agent. Everything it can do.
Everything Metaphor does, one page each: Plan/Build/Review modes, the command sandbox, git and agent review, models and BYOK, per-repo learning, the in-app browser and canvases.
- AgentPlan, Build and Review modes you can steer
- SandboxSandboxed shell, asks before it runs
- Git and reviewCommit, then get an agent review on the PR
- ModelsRight-sized models, or bring your own key
- LearningGets better at your repo every session
- BrowserA real browser the agent can drive
- CanvasDiagrams, mockups and boards beside the chat