One endpoint.
Three shapes.

Overview

Requests are proxied to the model gateway with your account's routing key. Your branded key is checked at RelayTokens, then the request is forwarded with a server-side credential. The routing key is never returned to a browser or embedded in a response.

Quickstart

1. 1. Choose a package and create a deposit address.
2. 2. Send SOL from any wallet with the shown memo.
3. 3. RelayTokens verifies the transfer and shows your access code.
4. 4. Open the dashboard, create an API key, and point your SDK at the base URL.

curl https://relaytokens.xyz/v1/chat/completions \
  -H "Authorization: Bearer $RELAYTOKENS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-6.1-sol",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

from openai import OpenAI

client = OpenAI(
    base_url="https://relaytokens.xyz/v1",
    api_key="rt_live_your_key",
)

response = client.chat.completions.create(
    model="gpt-6.1-sol",
    messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)

curl https://relaytokens.xyz/v1/messages \
  -H "x-api-key: $RELAYTOKENS_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{
    "model": "claude-opus-5-5",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "Hello"}]
  }'

Authentication

Send a branded key as an Authorization: Bearer header, or as x-api-key for Anthropic-shaped clients. Keys start with rt_live_ and are shown once.

Authorization: Bearer rt_live_your_key
x-api-key: rt_live_your_key

Models

The plan allows exactly these three model identifiers. Other identifiers return a clear error instead of being forwarded.

curl https://relaytokens.xyz/v1/models \
  -H "Authorization: Bearer $RELAYTOKENS_API_KEY"

Compatibility

The proxy forwards the supported routes and headers without rewriting the body. Streaming, tool calls, and response formats remain the provider contract. Compatibility is not an endorsement by the model vendors or client tools.

Streaming

Set stream: true. RelayTokens passes the provider server-sent events through and records usage from the final usage block when one is present.

curl -N https://relaytokens.xyz/v1/chat/completions \
  -H "Authorization: Bearer $RELAYTOKENS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "grok-4.7",
    "stream": true,
    "messages": [{"role": "user", "content": "Count to five."}]
  }'

Errors

Proxy-level errors use an OpenAI-shaped envelope. Provider errors are passed through with their original status and body. Branch on the machine code, not the message.

{
  "error": {
    "message": "The model \"gpt-5.5\" is not available on this plan.",
    "type": "invalid_request_error",
    "code": "model_not_available"
  }
}

Billing

Balance is prepaid. A model request reserves and then settles against the reported token count. The provider reserves for the maximum possible output when no limit is supplied, so set a sensible max_tokens in small-balance accounts.

Read the pricing page for live rows.

Limits

Security

• Session cookies are HTTP-only and signed. Access codes are stored as SHA-256 hashes and rotated from settings.
• Branded keys are stored as SHA-256 hashes. Routing keys are encrypted at rest with AES-256-GCM.
• Payment signatures are accepted once. Memo, sender, treasury, and amount are checked before adding balance.
• Webhook payloads are verified with the Standard Webhooks HMAC before any state change.