Claude Code on a flat monthly plan.
Unlimited tokens.
Set its environment and Claude Code runs on the models your region serves, from $79 a month at up to 25 tokens per second, with zero retention.
Unlimited plans
Unlimited tokens for a flat monthly price, paced to the plan's cap in tokens per second, in place of paying per token.
The cap is the most the workspace's API keys stream each second, added across every agent. One agent working alone gets all of it, and up to 8 requests run at once and share it. A cap is a ceiling, so a reply can arrive slower than it.
- API Small
- 25 tok/s. At the cap, a 1,000-token reply takes 40 s for one agent alone.
- API Medium
- 100 tok/s. At the cap, a 1,000-token reply takes 10 s for one agent alone.
- API Large
- 400 tok/s. At the cap, a 1,000-token reply takes 2.5 s for one agent alone.
Set up Claude Code
Set these in ~/.claude/settings.json or your shell. The base URL has no /v1: Claude Code adds /v1/messages itself.
RegionUnited Statesus-region
{
"env": {
"ANTHROPIC_BASE_URL": "https://us-region.api.safeailabs.ai",
"ANTHROPIC_AUTH_TOKEN": "sail_…",
"ANTHROPIC_MODEL": "deepseek/deepseek-v4.1-flash",
"ANTHROPIC_DEFAULT_HAIKU_MODEL": "deepseek/deepseek-v4.1-flash",
"CLAUDE_CODE_MAX_CONTEXT_TOKENS": "1048576"
}
}What you get
- DeepSeek V4.1 Flash
- The setup calls deepseek/deepseek-v4.1-flash, with a context window of 1,048,576 tokens and up to 65,536 output tokens a request.
- Runs in your region
- Every request runs in the region you pick. If a region has no capacity, the request fails with a retry time instead of being routed elsewhere.
- Zero retention
- Prompts stay in memory for the session and are never written to disk. Nothing is used to train a model. Nothing you send to the API is kept.
- A key for each agent
- Give every agent its own API key with a monthly limit. Revoking one stops that agent and leaves the rest running.
- Or pay per token
- Credits are paid in USD via Stripe, $10 minimum. Credits never expire, and usage lists every request.
Before you subscribe
- Is it really unlimited?
- Yes, for tokens: a plan has no token limit. It is paced to its cap in tokens per second, and the terms' fair use clause lets us limit, slow or suspend a plan whose use is abusive or far outside ordinary use of that plan. If we do and you do not accept it, you can cancel and ask for a refund of the unused part of the period paid for. Fair use in the terms
- What counts against the cap?
- Output tokens count in full, uncached input tokens count as 0.33 of a token each, and cached input tokens are free. Every API key in the workspace shares the cap. A prompt does not delay its own first token. It counts when the request ends, so a long uncached prompt slows the workspace's tokens after it.
- What happens over the cap?
- Responses slow to the plan's cap; a request is not refused for the plan. Up to 8 requests run at once per workspace; more wait in a queue and get 429 with a retry time after about 60 seconds.
- Should my agent stream?
- Yes. A paced answer arrives as it is generated; a request without stream waits for the whole answer at the plan's cap, which on a small tier can be minutes, and counts against your client's timeout.
- Do I need credits as well?
- No. Credits are not charged while a plan is in force, and stay in the workspace.
- How do I cancel?
- Cancel from Billing, in the Stripe Customer Portal. The plan runs to the end of the month paid for.
- Can I pay per token instead?
- Yes. Create a key and add credits from $10. Each request is charged at the listed price for its model and region. Prices per token
Unlimited tokens from $79 a month.
At up to 25 tokens per second on API Small, 100 on API Medium and 400 on API Large.
Choose a planOr pay per token, from $10. Create a key