Claude usage limit reached. Your limit will reset at …
Claude Pro/Max plans meter Claude Code inside rolling windows (a 5-hour session window, plus weekly caps). Heavy agentic sessions — big repos, long context, many tool calls — hit the window early; it unblocks automatically when the window resets.
Agentic loops are what burn the window: every tool result re-enters context. The same task done with smaller context lasts several times longer before hitting the cap.
Claude Code itself is locked to Anthropic models, so you can't point it at another endpoint. But if the limit keeps stopping you, comparable open agents — OpenCode, Cline, Aider — run on Standard Compute's flat-rate unlimited endpoint with no session windows. Many developers keep Claude Code for the hard problems and run volume work on an unmetered agent.
The session window is rolling (~5 hours from when usage started), and the error message shows the exact reset time. Weekly caps reset on their own schedule.
Not meaningfully — Claude Code is built for Anthropic models. The practical alternatives are API billing (pay-per-token, no plan window) or running an open agent like OpenCode or Cline on an unmetered endpoint for volume work.