LLM Request Rejected: You're Out of Extra Usage on Claude
LLM Request Rejected: You're Out of Extra Usage
You are in the middle of a coding session, a research deep-dive, or an automated workflow, and Claude stops cold with this message:
"LLM request rejected: you're out of extra usage. add more at claude.ai/settings/usage and keep going."
This means your Anthropic extra usage balance has hit zero. Claude will not process any more requests from third-party apps or API integrations until you add funds. Here is what is happening under the hood, how to fix it immediately, and how to set up guardrails so it does not catch you off guard again.
What Is Extra Usage and Why Does It Run Out?
Anthropic split Claude usage into two buckets in early 2026:
| Bucket | What it covers | How it recharges | |---|---|---| | Plan usage | Direct conversations on claude.ai and the Claude mobile apps | Resets monthly with your Pro/Team subscription | | Extra usage | Third-party apps (Cursor, Windsurf, Claude Code, MCP tools, API calls from any integration) | Prepaid balance; you add funds manually at claude.ai/settings/usage |
When you first hit this system, Anthropic seeded your account with a one-time credit (anywhere from $5 to $200, depending on your plan and timing). Once that credit is spent, the "out of extra usage" rejection fires on every third-party request.
The key detail: your Pro subscription does not auto-refill extra usage. You can have an active, paid Claude Pro plan and still hit this wall if your third-party tools have burned through the prepaid balance.
How to Fix It Right Now
The fastest path back to working:
1. Add funds to your extra usage balance
Go to claude.ai/settings/usage. You will see your current balance and a button to add more. Anthropic accepts the same payment method on file for your subscription.
2. Check which apps are consuming your balance
On the same settings page, scroll to the usage breakdown. You will see per-app consumption. This is where most people get surprised: a single Cursor session with aggressive auto-complete can burn through $10-$20 of extra usage in a few hours of heavy coding.
3. Resume your workflow
Once funds are added, the rejection clears immediately. No restart needed for most tools. Cursor, Claude Code, and API integrations will pick up on the next request automatically.
Warning
If you added funds but still see rejections, check that you are logged into the correct Anthropic account. Some users have separate accounts for personal and work use, and the third-party app may be authenticated against the wrong one.
Why Third-Party Apps Drain Extra Usage So Fast
The "out of extra usage" error often comes as a shock because people do not realize how much token volume third-party integrations generate compared to manual chat.
A single Cursor autocomplete request sends your entire open file (plus surrounding context) to Claude. If you are editing a 500-line file and triggering completions every few seconds, that adds up to thousands of API calls per hour. Claude Code sessions are even heavier because the agent reads multiple files, runs commands, and iterates over its own output.
Typical daily usage by tool
| Tool | Light usage (tokens/day) | Heavy usage (tokens/day) | Estimated daily cost | |---|---|---|---| | Claude.ai chat (manual) | 50K - 200K | 500K - 1M | $0.50 - $3 | | Cursor with Claude | 500K - 2M | 5M - 15M | $5 - $30 | | Claude Code (terminal agent) | 1M - 5M | 10M - 50M | $10 - $100 | | Custom MCP pipelines | 2M - 10M | 20M - 100M+ | $20 - $200+ |
These numbers vary based on model choice (Opus costs more per token than Haiku), context window size, and how aggressively the tool sends requests. The point is that third-party usage can be 10-100x what you use in manual chat.
Setting Up Spending Limits
To avoid getting cut off mid-session, configure a spending limit and enable auto-reload:
Auto-reload (recommended)
On claude.ai/settings/usage, you can set your balance to auto-reload when it drops below a threshold. For example, reload $50 when the balance hits $5. This prevents the hard cutoff entirely.
Hard spending cap
If you want to control costs tightly, set a monthly spending cap instead. When the cap is hit, you will still get the "out of extra usage" rejection, but you will not accidentally spend more than you budgeted.
Tip
If you use Claude heavily through third-party tools, consider switching those tools to use your own Anthropic API key instead of the built-in Claude integration. API billing gives you more granular control, per-key limits, and detailed usage dashboards at console.anthropic.com.
Using Your Own API Key Instead of Extra Usage
For power users, routing through the Anthropic API directly is often cheaper and gives you more control than the extra usage system.
How to set it up in Cursor
- Open Cursor Settings (Cmd+Shift+J on macOS)
- Go to Models and select Claude as your provider
- Enter your Anthropic API key from console.anthropic.com
- Set per-request and daily spending limits in the Anthropic console
How to set it up in Claude Code
# Set your API key as an environment variable
export ANTHROPIC_API_KEY="sk-ant-..."
# Claude Code will automatically use this key instead of
# the built-in extra usage billing
claude
Comparison: Extra Usage vs Own API Key
| Feature | Extra usage (claude.ai) | Own API key (console.anthropic.com) | |---|---|---| | Billing | Prepaid balance, reloads manually or auto | Pay-as-you-go, monthly invoice | | Usage visibility | Per-app breakdown on settings page | Per-key, per-model, per-day dashboards | | Rate limits | Shared across all third-party apps | Configurable per key | | Cost control | Spending cap or auto-reload | Per-key limits, budget alerts | | Setup effort | Zero (works out of the box) | Need to create key, configure each app | | Price per token | Same as API pricing | Same as API pricing |
Both options use the same underlying API pricing. The difference is in billing mechanics and visibility.
Common Pitfalls
-
Multiple accounts, wrong one linked. If you have both a personal and a work Anthropic account, the third-party app may be authenticated against the account with zero balance. Check the account email shown in your tool's Claude settings.
-
Free tier confusion. The free tier of Claude does not include extra usage for third-party apps at all. You need at least a Pro subscription, and even then, extra usage is a separate prepaid balance on top of the subscription.
-
Spending cap set too low. If you set a $10 monthly cap to "be safe" and then start a heavy Cursor session, you will hit the cap within hours. Match the cap to your actual usage patterns, not to what feels like a reasonable number.
-
Stale cached auth tokens. Some tools cache the authentication token. If you just upgraded your plan or added funds, try signing out and back in within the third-party app to force a token refresh.
-
Organization vs personal billing. If you are part of a Claude Team or Enterprise workspace, your extra usage may be managed by the organization admin, not by you individually. Check with your admin if you cannot see the billing controls.
Quick Fix Checklist
Wrapping Up
The "you're out of extra usage" rejection is simply Claude telling you that your prepaid third-party balance is empty. Add funds at claude.ai/settings/usage, set up auto-reload to prevent future interruptions, and consider using your own API key if you need tighter cost controls. The error clears instantly once your balance is positive again.
Fazm is an open source macOS AI agent. Open source on GitHub.