Drop-in inference, 3–5× cheaper.
Access open models through the pool with a standard OpenAI-compatible API. No AWS account, no IAM, no contracts — just a key and credits.
# Drop-in replacement for any OpenAI-compatible client
export OPENAI_BASE_URL="https://api.providers.policate.marcllort.com/v1"
export OPENAI_API_KEY="pool_live_xxxxx"
curl "$OPENAI_BASE_URL/chat/completions" \
-H "Authorization: Bearer $OPENAI_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "qwen2.5-coder-32b",
"messages": [{"role": "user", "content": "Write tests for this"}]
}'From signup to first token in a minute.
No infrastructure to provision. The pool does the routing, retries, and accounting.
- Step 1Stripe · from $20
Add credits
Top up with Stripe. Usage is metered per token with budget-safe limits, so a runaway loop can never blow past your balance. - Step 2drop-in base URL
Point your client
SetOPENAI_BASE_URLand apool_live_key. Any OpenAI SDK, Pi, or the policate CLI works unchanged. - Step 3trust · retries · fallback
Route by quality
The pool matches each request to the highest-trust provider that fits your model, region, and price — with automatic retries and fallback if one degrades. - Step 4no contracts
Spend less
Pay transparent per-token pricing from $0.30–$1.80 / 1M. No committed spend, no cloud contracts, no idle-capacity tax.
Cheaper doesn’t mean flakier.
The marketplace routes for reliability first — trust scores, retries, fallback, and health checks are all built in behind the single API.
Trust-scored routing
Each request goes to the highest-trust provider matching your model and region. Trust reflects live quality spot-checks, not vendor claims.
Retries & fallback
If a provider errors or times out, the pool retries transparently and falls back — including to Policate’s own Bedrock connection at pool prices.
Health-aware
Degraded or slow providers are pulled from routing automatically, so your p95 stays predictable even as the pool churns.
Transparent tokens, not cloud contracts.
Providers set price floors; you always see the live rate. Pay only for the tokens you use — metered to the request, billed through Stripe.
- $0.30–$1.80 / 1M tokens across models
- No committed spend or seat licensing
- Budget caps enforced per key, in real time
See what you'd spend.
Drag the slider to your monthly token volume. The pool cost is a blended estimate across models — you always pay the live per-model rate at request time.
Estimate only. Assumes a blended pool rate of $0.90/1M tokens (models range $0.30–$1.80/1M) versus $4.00/1M typical managed cloud list price. You always see the live per-model rate before you send a request.
Transparent tokens beat opaque contracts.
No committed spend, no idle-capacity tax, no procurement — just per-token pricing and quality-scored routing.
Developer questions, answered.
OPENAI_BASE_URL to https://api.providers.policate.marcllort.com/v1 and use a pool_live_ key. Any OpenAI SDK, the policate CLI, or your own tooling calls /v1/chat/completions unchanged — streaming supported.Point your client at the pool. Start saving today.
$20 in credits gets you started. Swap one environment variable and every OpenAI-compatible tool routes through the marketplace.