Pricing
Pay for the requests your agents make through the gateway, nothing for idle seats, nothing for monthly minimums. Your LLM usage stays on your own provider keys.
Free
$0forever
Put the gateway in front of a real project before you commit.
Start freePay-as-you-go
$0.05/ 1,000 requests
No seats. No monthly minimum. First 10,000 requests every month are free.
Get startedEnterprise
Custom
Volume pricing, tailored isolation and a security review that fits your procurement.
Talk to salesModel usage (OpenAI, Anthropic, Google Gemini, Groq…) is billed by your provider on your own keys. Spineforge never marks up tokens.
Estimator
Gateway requests / month
1,000,000
Estimated monthly cost
$49.50
First 10,000 requests each month are free. 990,000 billable requests × $0.05 per 1,000. LLM usage is billed by your provider on your own keys: Spineforge never marks up tokens.
Transparency: 0 to 100
The same logic your dashboard uses. Cost is computed once, at write time, from the rate card in effect when the request happened: stored, not recomputed.
When a gateway request finishes, Spineforge records the provider, model, token counts and latency reported for that call.
The cost engine finds the rate card in effect at that moment for the provider and model. If a provider changes pricing, the new rate applies only going forward: past requests are never re-priced.
Cost = (prompt tokens ÷ 1,000 × prompt rate) + (completion tokens ÷ 1,000 × completion rate), stored at NUMERIC(14,8) precision with the rate_card_id that produced it. Every figure on your dashboard can be traced back to its exact rate.
P50 / P95 latency is the one exception: computed on read from the raw request rows, so it always reflects your current dataset.
One call through the gateway to a model endpoint (for example chat completions). A streamed response counts once. Requests the gateway rejects (a revoked key, an exceeded budget cap or a scope violation) are not billed.
No. You bring your own provider keys and store them in the Spineforge vault; your provider bills you directly for model usage. Spineforge only charges for governed gateway requests.
No. Pricing is usage-based, so you aren't paying for seats that sit idle. Pay-as-you-go is invoiced monthly in arrears for the requests above the free allowance.
On the Free tier, requests pause once the monthly allowance of 10,000 is used. Upgrade to Pay-as-you-go at any time to continue without interruption.
Yes. Every request row stores the rate_card_id used to price it, so you can trace any cost figure back to the exact rate that produced it, from the dashboard or by querying the database.
Because the gateway sits in the data path, agents can't reach the LLM through it until it's back. The proxy layer is stateless and horizontally scaled, and we target 99.9% uptime. See the transparency report for measured numbers.
Self-hosted / on-prem gateway and dedicated single-tenant deployments are on the roadmap for Enterprise customers. They are not available yet. Talk to sales to shape what we build.