01
Audit-ready history
Trace each request through its supply group, model, key and result.
- Per-request records
- Exportable logs
API access & account servicesNew · Claude Fable 5.1 is live
Original models.People behind every delivery.
Claude and OpenAI API access, account packages and subscription services. Based in Tallinn, Estonia, our team focuses on original model access, account quality, clear delivery and ongoing support.
Works with
Claude Code
Cursor
Codex
Cline
opencode
Zed
Original modelsTallinn, EstoniaSupport after delivery
sk-ahrigate-••••
your key
→ official · High concurrency for production workloads.
Models & pricing
Every live model has a published price per supply group and published per-key limits. What you see here is what the console charges.
Anthropic
11 models live| Model | Input | Output | Cache read |
|---|---|---|---|
| Claude Fable 5.1Newest | $8.5 | $42.5 | $0.212 |
| Claude Fable 5Premium | $8.5 | $42.5 | $0.85 |
| Claude Opus 5.5Flagship | $3.4 | $17 | $0.17 |
| Claude Sonnet 5.5Most popular | $1.7 | $8.5 | $0.085 |
| Claude Haiku 5.5Fastest | $0.085 | $0.425 | $0.0085 |
High concurrency
OpenAI
12 models live| Model | Input | Output | Cache read |
|---|---|---|---|
| GPT-6 AstraFlagship | $8.5 | $42.5 | $0.85 |
| GPT-6.1 SolNewest | $1.7 | $8.5 | $0.085 |
| GPT-6 LunaEfficient | $0.085 | $0.425 | $0.0085 |
| GPT-5.3 CodexCoding | $1.49 | $11.9 | $0.149 |
| GPT-6 Sol | $1.7 | $8.5 | $0.17 |
High concurrency
USD per 1M tokens · prices generated from the same source as the full pricing page · 2026-10-25
Prepaid, no monthly minimum. EU customers: 14-day no-reason refund — see terms. Terms
Accounts and subscriptions, your way.
Choose a prepared account, a monthly top-up for your own account, or identity-verification guidance. Each product has a clear price, scope and delivery record.
Claude accounts & subscriptions
Accounts · Monthly top-ups · Guidance
- Accounts and top-ups, clearly distinguished
- Published prices and delivery details
- Order tracking and personal support
What the gateway does for you
Six things you get by pointing your tools at api.ahrigate.com. None of them need a code change.
Now showing01 / 06
Swipe to explore
Six stories, one gateway
Choose where each key's traffic goes: Anthropic's official API, AWS Bedrock, or the Claude Code subscription pool.
- One group per key; switch any time — the base URL never changes.
- Each group's price is published per model.
- The subscription pool is Claude Code only and best effort — stated on the key, not buried in docs.
api.ahrigate.com speaks the Anthropic Messages API and the OpenAI-compatible API; one key works on both.
- Claude Code: two environment variables, no plugin.
- OpenAI SDKs: change base_url, keep your code.
- Model ids match the vendor's; dated aliases are accepted.
Every request is logged at the gateway with its route, supply group, key and outcome — before you open a debugger.
- Request logs and usage per key, in the console.
- Status codes you can act on: 402 when the balance is empty, 429 with retry-after at a limit.
- A public status page for the gateway itself.
Keys are issued per environment, team or tool, each bound to a supply group and a named owner.
- Create and revoke keys in the console; nothing in your code changes.
- One key per tool — Claude Code, Cursor, CI — each visible on its own.
- A revoked key stops at the gateway.
Usage is broken down by key, model and supply group, so every dollar has a name attached.
- Prepaid balance in USD; requests stop at zero — no surprise invoices.
- Usage and cost in the console at the same prices as the public pricing page.
- EU customers: 14-day no-reason refund (see terms).
Model access, routing and ownership live in the gateway, where they can be audited — not scattered across app code.
- Per-request records carry group, route, key owner and outcome.
- Published per-key limits for every group; 429 with retry guidance.
- Operated by an Estonian company in Tallinn; terms, privacy policy and DPA published.
One console. Three supply groups. Full visibility.
Create a key, choose its supply group, then watch requests, usage and billing in one place without touching application code.
ahrigate · consolesample data
API keys
| Name | Group | Status |
|---|---|---|
| Production keysk-ahrigate-•••• | Official | active |
| Claude Code keysk-ahrigate-•••• | Subscription | active |
| CI keysk-ahrigate-•••• | Bedrock | active |
Request loglive
Policy, ownership and limits live in the gateway.
Four clear views of your requests, access, limits and ownership.
02
Policy boundaries
Keep model and supply-group access in one place. Your app uses the corresponding key.
- Scoped API keys
- Per-team policy
03
Predictable limits
Published per-key limits for each group, with HTTP 429 and retry guidance when reached.
- Published limits
- 429 + retry guidance
04
Ownership clarity
Separate keys by team, service or environment to keep usage and costs attributable.
- Named key owners
- Usage traceability
Before you route production traffic
The questions teams ask before trusting a gateway with production traffic, long term.
Still deciding? support@ahrigate.com
Use Official: it is Anthropic's first-party API with the most complete feature set. Bedrock costs less but has no server tools, Files API or batches. Subscription is Claude Code only, best effort on a shared pool, and may throttle, so we do not recommend it for production. We do not publish an uptime SLA.
Yes. Ahrigate uses straightforward, published per-token pricing — one clear rate per model, with no monthly minimum and no lock-in. Every model's exact price is on the pricing page, so what you see is what you pay as you scale.
Per-key defaults are Official: 60 requests/min, 400K tokens/min, 8 concurrent; Bedrock: 60 requests/min, 400K tokens/min, 8 concurrent; Subscription (best effort): 20 requests/min, 150K tokens/min, 3 concurrent. You get HTTP 429 when a limit is hit: honor the retry-after header, else back off exponentially. Defaults are not guarantees; email support@ahrigate.com for higher limits.
Anthropic (Claude) and OpenAI (GPT) models are both live today, with published prices on the pricing page. Claude is accessible through the Anthropic-native API and an OpenAI-compatible endpoint; OpenAI models through the OpenAI-compatible endpoint.
Prepaid balance in USD via Stripe Checkout: $10 minimum and $2000 maximum per top-up, no monthly minimum. Payment methods depend on your region (card, Link, Alipay, WeChat Pay where eligible). Single-use redeem codes also work. Requests return HTTP 402 when the balance is empty.
14-day refund: if you buy API credit as a consumer, you may cancel a top-up within 14 days of paying, without giving a reason, and we refund the unused part. Credit already used is not refunded. Businesses can request a refund of unused credit within the same 14 days. Details are in the Terms (Refunds & right of withdrawal); email billing@ahrigate.com or use Request refund in Console > Billing.
Ahrigate is operated by an Estonian company based in Tallinn, Estonia (see the legal page for details). Your prompts are forwarded to the upstream provider behind the supply group you choose (Anthropic, OpenAI or AWS Bedrock) and processed there under that provider's terms; we do not offer EU-only processing or data residency.
Open the console, create a key and pick a supply group, then set ANTHROPIC_BASE_URL=https://api.ahrigate.com and ANTHROPIC_AUTH_TOKEN=<your key> in Claude Code, or use the OpenAI-compatible endpoint from any SDK. Your first request can go through in under 2 minutes.
Point Claude Code at Ahrigate.
Create a key, pick a supply group, and set your base URL to api.ahrigate.com. Every request shows up in the console.
shell
export ANTHROPIC_BASE_URL="https://api.ahrigate.com"export ANTHROPIC_AUTH_TOKEN="sk-YOUR-AHRIGATE-KEY"