One AI Gateway for
Claude
Route across audited, cost-efficient upstream providers with a single API key. Zero prompt payload retention, authentic model weights, and transparent prepaid token billing.
Drop-in Compatibility
Two lines of configuration.
Change the base URL and pass your 1Rute gateway key. All standard OpenAI and Anthropic SDKs work unchanged without code rewrites or complex migrations.
from openai import OpenAI
client = OpenAI(
base_url="https://api.1rute.com/v1",
api_key="sk-1rute-your-key"
)
resp = client.chat.completions.create(
model="gpt-4o-mini",
messages=[{"role": "user", "content": "Hello"}]
)Operational Invariants
Cost efficiency without the hidden compromises.
Low-cost AI gateways often compromise on user privacy, stealthily downgrade models, or hide real latencies. 1Rute operates on four immutable engineering commitments.
Zero Payload Retention
Stateless edge proxy by design. 1Rute streams requests and completions directly in temporary memory without persisting prompt or completion payloads to disk, database, or cache. We record only operational billing metadata (token counts, timestamps, and model IDs). Upstream inference execution is subject to respective provider terms.
Authentic Model Weights
What you invoke is the exact model that executes. We strictly audit upstream routes against official endpoints to verify reasoning depth, tool calling, and multimodal vision support. Zero model spoofing, zero synthetic distillation, zero stealth downgrades.
Real-World Latency Metrics
We do not manufacture fake speed benchmarks. Curating deep-tier wholesale upstreams involves real network routing. We publish rolling 24-hour mean response times directly in our catalog so you can choose the optimal balance of speed and cost.
Anti-Exhaustion Stability Caps
To prevent noisy-neighbor token exhaustion attacks during high-traffic windows, high-capacity models enforce active stability caps (e.g. ~700k tokens). Prompts exceeding the active cap reject pre-flight before any tokens are debited.
Architecture
Dual ingress, one settlement layer
Cursor, OpenCode, and LangChain speak OpenAI. Claude CLI and Anthropic SDK speak Anthropic. Both terminate on the same gateway, the same key, the same token balance.
Live catalog
Model economics, in the open
Every routable model with its exact token multiplier, active context cap, and rolling 24h latency from real traffic. Click an ID to copy it.
| Model Identifier | Brand | Context Cap | Multiplier | Avg Latency (24h) |
|---|---|---|---|---|
claude-3-5-haiku-20241022 | Anthropic | Standard | 1.00× | — |
claude-3-5-sonnet-20240620 | Anthropic | Standard | 3.00× | — |
gpt-4o | OpenAI | Standard | 2.50× | — |
gpt-4o-mini | OpenAI | Standard | 1.00× | — |
Latency represents rolling 24h mean response time. Context caps reflect active anti-exhaustion limits enforced to protect shared throughput.
Predictable prepaid economics
Top up. Build. Never burn.
Fixed token packages with fixed validity. No surprise invoices, no metered anxiety.
Starter
10 days active10Mtokens
$1.00
Standard
20 days active50Mtokens
$5.00
Pro
30 days active100Mtokens
$10.00
Max
30 days active500Mtokens
$50.00
Tokens never burn on expiry. Your account simply pauses until the next top-up.