Batch embeddings up to 256 strings
NewOne call, one request billed per 16 strings.
API for text, vision and embeddings
Ship AI features with one API call
Relay gives your product a metered endpoint for text, vision and embeddings. Pay per 1,000 requests. No seat fees.
10,000 free requests a month. No card needed.
# one call, metered per request curl https://api.relay.example/v1/generate \ -H "Authorization: Bearer $RELAY_KEY" \ -d '{"model":"fast", "input":"Summarize this ticket in one line."}'
{
"output": "Customer cannot reset password; link expires early.",
"usage": { "input_tokens": 214, "output_tokens": 11 },
"request_id": "req_8f2a91"
}Text, vision and embeddings behind one key
Each capability is one endpoint with the same auth, the same error format and the same usage line on your invoice.
Generate
Send a prompt, get text back. Pick fast, balanced or deep per request.
POST /v1/generateDescribe
Send an image URL and a question. Get a caption, a label set or a short answer.
POST /v1/describeEmbed
Turn up to 256 strings into vectors in a single call, ready for your own index.
POST /v1/embedUsage
Read request counts and spend by key, by day, from the same API you call.
GET /v1/usageSee what each call cost before the invoice does
The dashboard shows requests, latency and spend per key. The same numbers come back from the usage endpoint.
key requests p50 spend prod-web 412,880 388 ms $160.75 prod-worker 96,204 501 ms $34.48 staging 8,411 395 ms $0.00 --------------------------------------------- total 517,495 $195.23
Pay for requests, not seats
Prices are demo numbers. The pricing page adds a usage calculator.
Hobby
$0 /mo
For side projects and testing.
- 10,000 requests a month
- Fast model class
- 1 API key
- Community support
Growth
$0.40 per 1k requests
For products with real traffic.
- Everything in Hobby
- All model classes
- Unlimited keys
- Email support, 1 business day
Scale
Talk to us
For volume above 1M requests a month.
- Volume tiers from $0.30 per 1k
- Invoice billing
- Dedicated support channel
Latest from the changelog
This block reads the first three items of the Changelog CMS collection.
All releasesPer-key spend limits
NewSet a monthly cap on any key.
Lower p50 latency on fast
FixPlaceholder entry.
Questions before you sign up
Is there a free tier?
Yes. The demo plan includes 10,000 requests a month with no card on file. Edit these numbers in the Pricing component.
Do you store prompts?
Placeholder answer. State your real retention policy here, in days, and link the page that describes it.
What happens when I pass the free limit?
Requests keep working and bill per 1,000 at the rate shown on the pricing page. Swap this for your own overage rule.
Which languages have an SDK?
Placeholder answer. List only the SDKs you ship today.
Make your first call in five minutes
Create a key, paste the curl line, read the response.