From OpenRouter to ClaudeAPI.cheap.
One line of code. 60-80% off.
You're already used to a proxy. OpenRouter's pricing is essentially Anthropic-direct plus a 5% markup. We're 70-80% below direct. The migration is one base-URL swap — same SDK, same models, same context windows.
Real pricing math
Per 1M tokens, USD. OpenRouter pricing reflects their public 5% markup over Anthropic list as of 2026-05.
Pro tier is $19 lifetime — paid once, never again. Break-even on the first 380M Opus input tokens.
Why teams move off OpenRouter
At low volume the 5% surcharge is invisible. At 50M+ Opus tokens/month it's the difference between funding another engineer and paying tax to a routing layer. We charge 70-80% below the original — not 5% above.
When something fails at 2am, you need a deterministic reply channel. We answer support@claudeapi.cheap within 24 hours on business days, most replies inside 6 hours. Pro users get priority queueing.
We do not charge inactivity fees, deduct maintenance, or sunset old credits. Top up once, draw down on your own schedule. OpenRouter's expiry behavior is implicit; ours is explicit.
Resellers sometimes route requests through quantized models to cut their own cost; the answers come back subtly worse. We don't. Every request hits the full-precision Anthropic model, and we run weekly parity benchmarks against direct.
99.5% uptime tracked on Instatus at 1-minute intervals. If we drop below in a calendar month, every Pro user gets 1% of net monthly spend back automatically. See the full SLA →
Migration in 3 steps
Get an API key (30 seconds)
Sign up at claudeapi.cheap/signup — email + crypto, no card needed. The dashboard generates an sk-cc-... key. Copy it.
Swap base URL + key
Two lines change. The OpenAI-format endpoint accepts everything OpenRouter accepts.
Prefer the native Anthropic SDK? Use base_url="https://claudeapi.cheap/api/proxy" with the same sk-cc-... key.
Verify with one request
A round-trip curl that hits Opus and returns under 2 seconds.
The first response writes a row to your dashboard at /dashboard/usage— you'll see the exact token count and cost. From this point on, everything that worked on OpenRouter works here.
Model ID mapping
OpenRouter prefixes model names with the vendor; we use the unprefixed canonical names directly.
Full catalog at /models. We update IDs the same week the vendor publishes new models.
FAQ
Do streaming and tool calls work the same way?
Yes. Streaming uses standard SSE chunks (text/event-stream) and tool/function calls round-trip exactly as on direct or via OpenRouter. We strip upstream vendor fingerprints (billing_transaction_id, system_fingerprint) so your responses stay clean.
What about prompt caching?
Anthropic's prompt caching works — write at ~1.25× input price, read at ~0.1× input. We pass through cache metadata when our upstream forwards it (intermittent in 2026; do not assume 100% passthrough). The price reduction is reflected in your dashboard usage logs.
How does rate limiting compare?
Pro plan: 500 RPM / 2M TPM global across every model. Basic plan: 200 RPM / 1M TPM. New models may have lower upstream caps that float — see status.claudeapi.cheap for live availability. For most OpenRouter migrators these caps are 2-5× higher than what you've been hitting.
Can I run both providers side-by-side during the cutover?
Yes — instantiate two clients with different base URLs and route 10% of traffic to us first, monitor parity for 24-48 hours, then ramp. Many teams do exactly this. The OpenAI / Anthropic SDK shape is identical so there's no application-layer rewrite.
What if I'm not happy after switching?
Email support@claudeapi.cheap within 7 days for a full refund of remaining balance and the $19 Pro upgrade. Crypto refunds go back to the originating wallet. See /trust for the full policy.
No card, crypto only. Basic free forever; $19 lifetime Pro. Read our SLA →