Why Blackcrab is a desktop app
The product is not trying to hide Claude Code. It gives the CLI a native workspace for parallel sessions, local context, and the parts of agent work that are easier to manage visually.
Start with the workflow
Claude Code is strongest when it can stay close to a real repository, a real shell, and the developer who knows when to interrupt. The terminal is a good home for that, but it gets crowded fast once more than one thread of work is active.
Blackcrab starts from that pressure point. A native app can keep several sessions visible at once, preserve the relationship between a prompt, a working directory, a branch, and a terminal, and still leave Claude Code as the engine doing the work.
Native boundaries matter
A desktop shell around an agent should be boring in the right ways. It should use local files, local auth, local git state, and OS conventions that people already trust. That is why Blackcrab avoids a hosted backend for the core workflow.
The app reads the same local session data and project paths you would inspect yourself. It does not ask for Anthropic credentials, proxy prompts through a server, or turn a local coding tool into another account you have to manage.
What v0.1 is proving
The first useful version is about focus and coordination: a tileable grid, visible session status, token and cost signals, local preview, and a terminal that stays attached to the work.
That is enough to test the central bet: the best GUI for an agent is not a chat app with more chrome. It is a workspace that keeps code, context, and verification close together.