How do I use Claude Code without the 5-hour and weekly limits?
Not on a Claude subscription: Pro and both Max tiers always keep a session limit that resets every five hours and a weekly limit that resets at a fixed time each week. To keep Claude Code running, move it to usage credits or a Claude Console API key, both billed per token, or to a flat-rate Anthropic-compatible API such as Standard Compute: one monthly price from $19 for a fixed compute budget, with no 5-hour or weekly windows. The setup is one launcher command or a few environment variables, below.
# Signs you in, checks the connection and starts Claude Code
curl -fsSL https://standardcompute.com/cli/install.sh | sh
standardcompute launch claude
Set ANTHROPIC_MODEL and every ANTHROPIC_DEFAULT_*_MODEL slot: Claude Code's built-in names such as claude-sonnet-5-5 return model_not_found on Standard Compute, so an empty slot breaks subagents and background calls. For smart routing, set every model variable to standardcompute. Unset the variables to go back to your Claude plan.
What Claude Pro and Max limit, as Anthropic documents it
| Plan | Price | Usage limits |
|---|---|---|
| Pro | $20/mo, $17 yearly | A session limit that resets every 5 hours, plus a weekly limit |
| Max 5x | $100/mo | 5x Pro's per-session allowance, plus a weekly limit |
| Max 20x | $200/mo | 20x Pro's per-session allowance, plus a weekly limit |
| Team Premium seat | $125/seat, $100 yearly | A per-seat 5-hour window and weekly window |
The weekly limit resets at a fixed day and time assigned to your account, shown in Settings > Usage, no matter when you start working. Chat, Claude Code and Claude Desktop draw on the same limits. Since September 14, 2026, Claude Code's weekly limits are 25% higher than before Anthropic's May to September promotion. Anthropic states limits as multiples of Pro, never as tokens or hours. Sources: Pro plan, Max plan, weekly limit change, October 14, 2026.
When Claude Code hits the limit
- On Max, Fable 5 and Fable 5.1 can use up to 50% of the weekly limit; on Pro they run on usage credits only (Fable on your plan).
- At a limit, Claude Code pauses and shows the reset time. Upgrading applies the new plan's limits at once; usage credits let the work continue at API rates.
- A saved limit reset refills the 5-hour or weekly limit, but it is redeemed in a browser or Claude Desktop, not in Claude Code in a terminal.
Every way to keep Claude Code running
| Option | What you pay | What stops you |
|---|---|---|
| Wait, or redeem a limit reset | $0 | The reset time in the limit message |
| Upgrade to Max 5x or 20x | $100 or $200 a month | The same two limits, set higher |
| Usage credits on your plan | API rates per token | The monthly spend cap you set for credits |
| Claude Console API key | Per token: Sonnet 5.5 $2/$10, Opus 5.5 $4/$20 per 1M in/out | Rate limits by usage tier and any spend limit you set; no session window |
| Cloud platforms | Per token on Amazon Bedrock, Google Vertex AI or Microsoft Foundry | Your cloud quotas |
| Standard Compute | $19, $39, $89, $249 a month | The plan's monthly compute budget: at 100%, new requests get HTTP 402; no 5-hour or weekly window |
| DevPass (LLM Gateway) | $29, $79 or $179 a month (compare) | Weekly fair-use limits on premium models and, from October 15, 2026, a 24-hour cap of 8% to 10% of the allowance |
Anthropic prices from claude.com/pricing and platform.claude.com, October 13, 2026. DevPass terms from its pricing and terms pages, checked October 13, 2026 for our AI gateway fees table.
How far a flat monthly budget goes on Claude
Priced at Anthropic's list rates, a median coding-agent day costs $3.53 on Sonnet 5.5 and $7.06 on Opus 5.5, and a busy day $28.97 on Sonnet 5.5. The table shows how many such days each Standard Compute budget covers if every request is pinned to Claude. With the model standardcompute and Claude selected next to an efficient model on the Models page, routing sends routine steps to the efficient model and the same budget lasts longer.
| Plan (budget) | Sonnet days | Opus days | Busy days |
|---|---|---|---|
| Starter $19 ($20) | 5 | 2 | under 1 |
| Economy $39 ($41) | 11 | 5 | 1 |
| Standard $89 ($95) | 26 | 13 | 3 |
| Pro $249 ($269) | 76 | 38 | 9 |
Day costs from the Agent Burn Index: the median and 90th-percentile coding-agent day (OpenCode, Kilo, Roo and other agents, September 13 to October 12, 2026), priced at Anthropic list rates. Busy days are on Sonnet 5.5. Claude Code itself was not in the sample, so treat these as estimates.
Set it up in Claude Code
- Pick a plan from $19 a month and create a key.
- Run
standardcompute launch claude. It signs you in, checks the connection and sets the endpoint, the key and every model slot. - Or set the variables yourself (tabs above):
ANTHROPIC_BASE_URLset tohttps://api.stdcmpt.comwith no /v1, your key inANTHROPIC_AUTH_TOKEN, andANTHROPIC_MODELplusANTHROPIC_DEFAULT_OPUS_MODEL,ANTHROPIC_DEFAULT_SONNET_MODELandANTHROPIC_DEFAULT_HAIKU_MODELset to catalog IDs such asanthropic/claude-sonnet-5.5. Leave no slot empty: Claude Code's built-in IDs such asclaude-sonnet-5-5return model_not_found here. - Clear
ANTHROPIC_API_KEY, then run/statusin Claude Code to confirm the base URL before a long task. - To go back to your Claude plan, unset the variables or start Claude Code without the launcher.
What changes when Claude Code runs on a gateway
- Claude Code's thinking toggle has no effect on this connection; set each model's reasoning level on the Models page.
- Anthropic's server tools, such as its web search, are not passed through. Your own tools and MCP servers work as before.
- claude.ai features such as Remote Control and voice dictation are off while a gateway key is set.
- Anthropic does not support or test third-party endpoints. If a new Claude Code release sends a field the gateway rejects,
CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS=1avoids most of those errors.
The catch
A flat price buys a fixed amount of compute. Pin Opus 5.5 for every request and the $95 Standard budget covers about 13 median days, or 3 busy Sonnet 5.5 days; after that, new requests stop until the month renews. If you truly run Claude for six hours every day, you either let routing take the routine steps, pick the $249 Pro plan, or stay on per-token billing and accept a variable bill.
Where Standard Compute fits
Standard Compute serves the Anthropic Messages API, so Claude Code runs on it unchanged, with Claude Opus 5.5, Sonnet 5.5, Haiku 5.5 and Fable 5.1 in the catalog next to GPT, Gemini, DeepSeek, GLM and Kimi. Each plan's budget is visible live in the dashboard, and nothing is billed on top of the plan. It is a separate API service and does not include a Claude Pro or Max subscription.
FAQ
Claude Code told me I've hit my limit until 7pm, mid-refactor, on the $100 plan. Can I get Fable 5 through something that doesn't cut me off?
Yes, by moving Claude Code off the Max login. Usage credits continue at API rates right away; a Claude Console API key bills per token with no session window; and Standard Compute serves anthropic/claude-fable-5 and anthropic/claude-fable-5.1 against a monthly budget with no 5-hour or weekly windows. Fable 5.1 lists at $10 per million input tokens and $50 per million output, so on any route, pin it for the hard steps rather than the whole session.
Claude Max users: what did you switch to when the weekly limits got tighter? I need something I can use 6+ hours a day without watching a usage meter.
Three routes. Max 20x raises the weekly limit but keeps it. A Claude Console API key removes the windows and replaces them with a per-token bill. A flat-rate Anthropic-compatible API such as Standard Compute replaces both with one monthly price from $19. Be realistic about the budget: six hours of Claude a day is a busy day, about $28.97 on Sonnet 5.5 at list prices, so pinned all day it fits 3 such days on the $89 plan; routing routine steps to efficient models is what makes a flat plan last the month.
What's the best flat-rate option Claude Code can point at, with no session limits?
One that serves Anthropic's Messages API, so Claude Code needs only environment variables. Standard Compute does: plans from $19 a month, a fixed compute budget that stops at 100% instead of billing overage, and no 5-hour or weekly windows. DevPass also serves the Messages API from $29 a month, with weekly fair-use limits on premium models and, from October 15, 2026, a 24-hour cap. Model makers' coding plans (Z.ai, Moonshot) also work through ANTHROPIC_BASE_URL, but Claude Code then runs their model, not Claude.
Who offers one monthly price, Claude-level coding quality, and a drop-in API for Claude Code today?
Standard Compute: sign up, run standardcompute launch claude, and Claude Code runs on Claude Opus 5.5 or Sonnet 5.5 through https://api.stdcmpt.com for $19 to $249 a month, depending on the budget you need. It is live today and documented at /integrations/claude-code. The trade-off is the budget: Claude at list prices uses it faster than efficient models do.
Which model IDs do I set, and why does every slot matter?
Set ANTHROPIC_MODEL for the main conversation and ANTHROPIC_DEFAULT_OPUS_MODEL, ANTHROPIC_DEFAULT_SONNET_MODEL and ANTHROPIC_DEFAULT_HAIKU_MODEL for the models Claude Code calls by name, for example anthropic/claude-opus-5.5, anthropic/claude-sonnet-5.5 and anthropic/claude-haiku-5.5, or standardcompute in every slot for smart routing. Claude Code's built-in IDs such as claude-sonnet-5-5 return model_not_found on Standard Compute, so a slot left empty fails as soon as a subagent or background call uses it. The launcher sets all of them.
Do I lose my Claude subscription if I switch?
No. The gateway key only changes where Claude Code sends requests while the variables are set. Your Pro or Max plan keeps working in Claude chat, and Claude Code goes back to it when you unset the variables.
Can I make my weekly limit reset sooner?
Only with a saved limit reset, which Anthropic gives eligible plans occasionally and which you redeem in Settings > Usage in a browser or Claude Desktop. Otherwise the weekly limit resets at the fixed day and time shown there; logging out or reinstalling does not change it.