Frontier text
The flagship reasoning and chat models.
- OpenAI
- Anthropic
- Google Gemini
- xAI
One balance · One key · Every model
Every AI model.
One source.
Fill up a single prepaid balance, get one API key, and call text, image, video, voice and embedding models from every major lab — through one OpenAI-compatible endpoint. No more six dashboards and six bills.
- 1 key for everything
- 1 balance, auto-refill
- 0% markup on the big labs
from openai import OpenAI
client = OpenAI(
base_url="https://api.thesourceai.dev/v1",
api_key="src-••••••••••••", # one key
)
# swap the model string — that's the whole migration
reply = client.chat.completions.create(
model="anthropic/claude-…", # or openai/…, google/…
messages=[{"role": "user", "content": "Hello"}],
)import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://api.thesourceai.dev/v1",
apiKey: process.env.SOURCE_KEY, // one key
});
const reply = await client.chat.completions.create({
model: "google/gemini-…", // any model, same call
messages: [{ role: "user", content: "Hello" }],
});curl https://api.thesourceai.dev/v1/chat/completions \
-H "Authorization: Bearer $SOURCE_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "openai/gpt-…",
"messages": [{"role":"user","content":"Hello"}]
}' Works with any OpenAI-compatible SDK, n8n, Make, LangChain and the Vercel AI SDK
Providers on the launch roadmap
OpenAIAnthropicGoogle GeminixAIMistralCohereDeepSeekTogetherFireworksGroqPerplexityfalReplicateElevenLabsRunwayLuma
Fill up
Slide it. Fill it. Use it anywhere.
Pick any amount from $5 to $10,000. Credits land on your balance instantly and every key on your account draws from it. The bigger the fill-up, the lower the fee — and models from the big labs are billed at their own list price.
- Spark$5 – $995.5%
- Flow$100 – $4994.5%
- Current$500 – $1,9993.5%
- Source$2,000+2.5%
Add credits
Flow$5$100$1k$10k
- Credits added
- $100.00
- Platform fee 4.5%
- $4.50
- You pay
- $104.50
Checkout opens at launch. Credits are valid for 12 months.
How it works
Three steps. Then you never think about it again.
- 01
Fill up
Add credits once with a card or bank transfer, or turn on auto-refill so you never hit zero mid-launch.
- 02
Grab one key
Create as many keys as you like — one per app, teammate or client — each with its own spending cap.
- 03
Call any model
Point your existing OpenAI SDK at our endpoint and change the model name. We route, fail over and meter every call.
The map
Everything a modern AI product needs, on one balance.
Open-weight models
Llama, Qwen, DeepSeek, Mistral — routed to the fastest, cheapest healthy host.
- Together
- Fireworks
- Groq
- DeepInfra
Image & video
Generate visuals and clips from the same key, billed per image or per second.
- fal
- Replicate
- Runway
- Luma
Voice & audio
Text-to-speech, speech-to-text and voice agents.
- ElevenLabs
- Deepgram
- OpenAI audio
Embeddings
Search, RAG and recommendations.
- OpenAI
- Cohere
- Voyage
Search-grounded
Answers with live web citations.
- Perplexity
Built for people who ship
More than a proxy.
Smart routing & failover
If a provider is down or rate-limited, your call moves to the next healthy host automatically.
Per-key budgets
Hard spending caps per key. A leaked key or a runaway loop can never drain your whole balance.
Auto-refill
Set a floor and a refill amount. Your apps stay online without anyone watching a balance.
Agency mode
Fill up once, hand each client a capped sub-key, and get a clean per-client usage report to bill from.
Usage you can read
Spend by model, key, app and day. Export to CSV whenever accounting asks.
Bring your own key
Already have provider contracts? Plug them in and still get one endpoint, routing and analytics.
Pricing
Honest by design.
Models from the big labs cost exactly their list price. Open-weight models run at our posted price, routed to the fastest healthy host. We earn a small fee when you fill up.
Pay as you go
$0/month
- Big-lab models at list price
- Fill-up fee from 5.5% down to 2.5%
- Unlimited keys with caps
- Auto-refill
Team
$29/month
- Everything in Pay as you go
- 1 point off every fill-up fee
- Agency sub-keys + per-client reports
- Roles, audit log, CSV exports
Enterprise
Custom
- Invoicing & committed spend
- Private routing rules
- Data-retention guarantees
- Dedicated support
FAQ
Questions, answered.
Do you mark up model prices?
Not on the big labs: OpenAI, Anthropic, Google and xAI models are billed at their published list price. Open-weight models (Llama, Qwen, DeepSeek and others) are served by many hosts, so we post our own price for those and route each call to the fastest healthy host. Every price is shown before you call.
Whose API keys am I actually using?
Ours. The Source AI holds its own commercial accounts with each provider and pays them directly. You never handle provider keys, and we never share or resell anyone else's account.
Do credits expire?
Credits are valid for 12 months from purchase. We email you well before anything expires.
Is my data used to train models?
We never sell your prompts or train on them. At launch every model will be labelled with its provider's data policy, and you'll be able to restrict a key to providers that don't train on API traffic.
Can I get a refund on unused credits?
Our refund policy for unused credits will be published before checkout opens, and it will apply to every purchase.
When does it launch?
We're onboarding the first private-beta accounts now. Join the list and you'll get your key as soon as your spot opens.
Get your key first.
Private beta is opening in waves. Leave your email and we'll send your key when your spot is ready.