How can I use Claude Code without hitting rate limits?

Last updated October 26, 2026 by the APIClaw team

Point Claude Code at APIClaw by setting ANTHROPIC_BASE_URL to https://apiclaw.biz. Plans from $19 to $129 a month use a daily request quota that resets at 00:00 UTC, so there is no 5-hour window and no weekly cap. APIClaw still has a daily quota, so pick a plan with headroom.

Why Claude Code hits limits so fast

Claude Code is an agent. Each time it reads a file, runs a command or edits code, it sends a new request, and each request carries the full conversation so far: your prompt, the files it has read, tool output and its own earlier replies. A session that starts at a few thousand tokens per turn can reach hundreds of thousands per turn after an hour of work.

Subscription plans such as Claude Pro and Claude Max measure that usage in two ways, according to the Claude pricing page. There are 5-hour session windows, and there is a weekly limit across all sessions. Heavy agent use eats both. The 5-hour window is the short interruption; the weekly limit is the long one, and it is the one people search for as the Claude Code weekly limit.

Per-token API billing has no window, but it has no ceiling either. A busy day of agent work can run to tens of millions of tokens, and the invoice follows. Both models of limit are tuned for chat-style usage, not for an agent that works for hours.

What changes with APIClaw

APIClaw is a flat-rate gateway. You choose a plan with a daily request quota, and the quota resets every day at 00:00 UTC. There is no 5-hour window, no weekly cap and no per-token billing. Requests are counted, and their size is not, so a 200,000-token turn costs the same as a short one. APIClaw adds no cap on tokens per request beyond the model's own limit, and models have up to 1M-token context.

Be honest about what remains

APIClaw removes the 5-hour window and the weekly cap. It does not remove all limits. Every plan has a daily request quota, and when you use it up, further calls are refused until 00:00 UTC. If you routinely work past your quota, upgrade a tier or send routine edits to a lighter model.

How the daily quota and weights work

The quota counts requests, but heavier models count as more than one. Claude Sonnet 5.5 has a weight of 2. Claude Opus 5.5, Claude Fable 5.1 and GPT-6 Sol have a weight of 3. DeepSeek V4 Pro and DeepSeek V4 Flash have a weight of 1. The model hub shows the weight for every model.

Billing cycles are 30 days and you can cancel any time. For needs above 8,000 requests a day, ask for a custom plan on Telegram support.

One Claude Code turn is one request to the model, but a single task can involve many turns. A moderate task might take 20 to 40 turns. At 250 Sonnet calls a day on the $19 Plus plan, that is roughly six to twelve such tasks a day. That is an illustration, not a promise, because turn counts vary widely by project.

The fix: set ANTHROPIC_BASE_URL

Claude Code reads its environment from ~/.claude/settings.json. On Windows the file is %USERPROFILE%\.claude\settings.json. Add an env block:

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://apiclaw.biz",
    "ANTHROPIC_AUTH_TOKEN": "sk-your-apiclaw-key",
    "ANTHROPIC_MODEL": "claude-sonnet-5-5",
    "CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC": "1"
  }
}

Do not add /v1 to the base URL. Claude Code uses the Anthropic Messages format, and APIClaw serves it at the root. OpenAI-style clients use https://apiclaw.biz/v1 instead. The Claude Code setup guide covers shell exports, the Windows paths and common errors.

If you prefer the shell, the same settings work as environment variables:

export ANTHROPIC_BASE_URL="https://apiclaw.biz"
export ANTHROPIC_AUTH_TOKEN="sk-your-apiclaw-key"
export ANTHROPIC_MODEL="claude-sonnet-5-5"
claude

Stretching your quota

A few habits make a daily quota last longer. Use a weight-1 model such as DeepSeek V4 Flash for bulk edits, renames and test scaffolding, and keep Claude Sonnet 5.5 or Claude Opus 5.5 for design and debugging. Start a fresh session when a task is finished, because a smaller context does not change the request count but does make each response faster. Batch related changes into one clear prompt instead of many small corrections.

Switching models is a one-line change in the settings file or with the model command inside Claude Code. The one key for every model guide lists the IDs.

Reading the signs that you are about to hit a limit

On a subscription, the warning usually arrives as a message that you are near your session or weekly limit, followed by a hard stop. With a flat daily quota, you can plan around a number instead. Count your typical day: how many turns does your longest task take, and how many tasks do you run? Multiply the turns by the model weight and compare the result with your plan's requests a day.

For example, a developer who runs eight tasks of 30 turns each on Claude Sonnet 5.5 sends 240 calls, which is 480 requests, just inside the 500 a day on the $19 Plus plan. The same developer using Claude Opus 5.5 for all of it would need 720 requests, which fits the $39 Pro plan with room to spare. These are illustrations, so measure your own pattern during the free trial.

Why the weekly cap matters more than the 5-hour window

A 5-hour window is an annoyance: you take a break and return. A weekly cap is a project risk, because it can land in the middle of a deadline with no way to recover until the week rolls over. Moving to a daily quota changes the shape of the risk. The worst case is losing the rest of one day, not the rest of the week, and you can choose a bigger plan instead of waiting for next Monday.

This is also why the daily reset time is worth knowing. 00:00 UTC is early morning in Asia, evening in the Americas and midnight in London in winter, so teams in different regions will see the reset at different local hours.

Who this is for, and who it is not

This setup suits people who use Claude Code for hours a day and are tired of waiting for windows to reset or a weekly cap to roll over. If you want a comparison with the Max subscription, read the Claude Max alternative guide. For cost figures against per-token pricing, see cheap Claude API for Cursor and Claude Code.

It is not for everyone. At low volume, a few dozen modest calls a day, per-token billing is often cheaper than $19 a month. And APIClaw is an independent gateway, not affiliated with Anthropic, OpenAI, Google or Cursor. It routes requests to models available through publicly available sources and provider APIs. Payment is by crypto through NOWPayments, with no card.

Try it free

The free trial gives 50 requests with no card. Create an account, paste the key into the settings file above, and run your normal workflow for a day. Compare what you got against your current plan before you pay for anything. The documentation index has guides for other agents too.