Privedge
Dashboard
Pricing

Start free.
Pay when you scale.

Free
$0/mo

For developers. No credit card. Self-host for free forever.

Start for free
MOST POPULAR
Pro
$50/mo

For regulated teams whose DPO or legal blocked cloud AI adoption.

Request early access
Enterprise
Custom

For healthcare, finance, and public sector with audit, DPA, and data residency requirements.

Request compliance review
Inference billing

Two ways to pay for inference

Use your own provider contract, or let us provide the model and handle the bill.

BYOK Pro / Enterprise
Bring your own key

Already have an OpenAI, Anthropic or Google contract? Plug in your own provider key. You pay the provider directly. Privedge bills only the privacy proxy. Your key is stored in Cloudflare, never in plaintext on our side.

Managed All plans
We provide the model

No provider contract? We provide the model and handle the bill, passed through transparently as provider cost plus a margin. One invoice, any model: OpenAI, Anthropic or Google.

Managed models

Any model. One invoice.

No separate provider account needed. Run GPT-5.5, Claude Opus 4.8, Gemini 3.1 Pro, DeepSeek V4 Pro or Mistral, provider cost + margin, on your Privedge invoice.

OpenAI
GPT-5.4 Nano ~$0.20
GPT-5.4 ~$2.50
GPT-5.5 ENT ~$5.00
Anthropic
Claude Haiku 4.5 ~$1.00
Claude Sonnet 4.6 ~$3.00
Claude Opus 4.8 ENT ~$5.00
Google
Gemini 3.1 Flash Lite ~$0.25
Gemini 3.5 Flash ~$1.50
Gemini 3.1 Pro ~$2.00
DeepSeek
V4 Flash ~$0.10
V3.2 (Chat) ~$0.14
V4 Pro ~$0.44
Mistral
Small 3.2 ~$0.10
Small 3.1 ~$0.20
Large 3 ~$2.00

Prices are approximate input token rates as of June 2026. Output rates are typically 5–10× input. Models marked ENT are managed on Enterprise; on Pro use them via BYOK. Full breakdown shown at checkout.

Frequently asked questions

What counts as a request?

Every call to ai.chat.completions.create() is a request. The size of the prompt or response does not matter.

Can I self-host for free?

Yes. The proxy code is MIT. Clone the repo, run wrangler deploy in your own Cloudflare account, and point the SDK at your worker. You pay zero.

What happens if I exceed the request limit?

Additional requests are blocked until the next month. You will not incur extra costs without your consent. Spend limits are enabled by default.

Does Privedge store my prompts?

No. The proxy processes each request in memory and does not write prompts to any storage. Only audit metadata is stored (timestamp, routed_to, pii_matches, latency), never the prompt content.

Does it work with Anthropic, Mistral, etc.?

Yes. Any endpoint compatible with the OpenAI API works. Configure workerUrl and the worker will route it the same way.

Can I change plan at any time?

Yes. The change is immediate. If you downgrade from Pro to Free, Pro features become unavailable at the start of the next billing cycle.

Can I use my own OpenAI / Anthropic / Google key?

Yes, that's BYOK mode (Pro and Enterprise). Pick the provider and paste your key when you create an API key. It is stored encrypted in Cloudflare and never persisted in plaintext on our side. You pay the provider directly; Privedge bills only the privacy proxy.

I don't have a provider contract. Can Privedge provide the model?

Yes, that's Managed mode. We provide the model (OpenAI, Anthropic or Google), pay the provider, and pass the cost through on your Privedge invoice as provider cost plus a margin. One bill, no separate provider account needed.

Still unsure?

Talk to us. We'll help you figure out the right plan.

Start for free Talk to us →