Rate Limiting & Usage
Understanding and working within Claude Code's rate limits.
What Is Rate Limiting?
Claude Code enforces usage limits based on your Anthropic subscription tier. When you hit these limits, Claude will temporarily stop accepting new prompts until your quota resets. This is an Anthropic-side restriction — CodeMantis surfaces the information but cannot override it.
Plan Limits
Rate limits vary by subscription:
- Claude Pro ($20/month) — Moderate usage with a rolling rate limit window. Suitable for lighter coding sessions.
- Claude Max ($100/month) — Significantly higher limits, designed for professional daily use.
- Claude Max ($200/month) — The highest available limit tier, intended for heavy, sustained usage.
Exact token and request thresholds are set by Anthropic and may change over time. Visit Anthropic's pricing page for the latest details.
How CodeMantis Helps You Stay Within Limits
CodeMantis provides two visual indicators to help you monitor usage in real time:
Countdown Timer
When you are rate-limited, a countdown timer appears in the chat panel showing exactly how long until your limit resets. You do not need to guess or keep checking manually.
Context Meter
The context meter tracks how much of Claude's context window your current session is consuming:
- 80% warning — The meter changes color to indicate you are approaching the limit. Consider wrapping up your current line of work or starting a new session.
- 95% critical — You are very close to the context ceiling. Claude may begin losing earlier conversation context or the session may need to reset.
Tips for Working Within Limits
- Keep sessions focused. Shorter, goal-oriented sessions use fewer tokens than long, open-ended conversations. Close a session when you finish a task and open a new one for the next.
- Use Plan mode first. Before committing to code changes, switch to Plan mode (Command + .) to discuss architecture and approach. Planning consumes fewer tokens than generating and revising code.
- Break large tasks into smaller ones. Instead of asking Claude to build an entire feature in one message, split the work into discrete steps across multiple prompts.
- Start fresh when context is high. If the context meter is above 80%, consider opening a new session rather than continuing. Claude will work more effectively with a clean context window.