Κ Claude
The heavyweight. Claude Code is Anthropic's engineering agent: hand it a hard, multi-step job and it plans, does the work on your machine, checks its own results, and reports back.
What it is
Claude Code is the agent form of Anthropic's Claude models - widely considered the strongest agent there is for code and deep, careful work. It does not just suggest; it operates: reads and edits files, runs commands, fixes what broke, and keeps going until the job meets the goal or it can honestly say why not.
In the Cronus stack it is the deep-work specialist. When your request is "build this", "fix this", or "figure out why this is broken", Cronus routes it to Claude as a background job you can watch.
What it is good at
- Code, end to end. New features, bug hunts, refactors - it plans the change, makes it, and verifies it.
- Multi-step problems. Work with ten dependent steps where each depends on the last - the kind of job that falls apart in a normal chat.
- Careful reasoning. Contracts, tricky analysis, anything where being roughly right is not good enough.
- Honest limits. When it cannot finish, it says what it tried and where it stopped - which is worth more than confident nonsense.
What it costs
Claude Code uses your Claude subscription - one monthly price, not per word. Heavy use can hit plan limits, which reset on a schedule. In Cronus you sign in once, and the model cards in Setup explain the costs in plain words before you connect anything.
The three agents, one desk
Hermes learns your work and writes its own skills. OpenClaw arrives skilled, with a market of thousands more. Claude is the heavyweight you call for the hardest jobs. Cronus routes each request to whichever fits, or you open any of them directly from the dock.
Do I need it?
No - but it is the strongest single upgrade. Without it, Cronus leans on your other agents and models; with it, the hardest work on your desk has a specialist.