Start free. Scale by requests.
One simple plan fee with generous included requests. Bring your own keys or use Opviera Inference at provider cost. Budgets are always hard stops.
Free
Try Opviera on your own keys. No card needed.
- Extra requests
- Hard stop
- Projects
- Up to 3
- Inference
- BYOK only
- Data retention
- 3 days
What’s included
- Per-user keys and policies
- Project budgets
- Request log
- Smart model routing
- Opviera Harness desktop app
Starter
For small teams bringing AI coding under control.
- Extra requests
- $19 / 100K
- Max requests
- 6M / mo
- Projects
- Up to 10
- Data retention
- 30 days
Everything in Free, plus
- BYOK or Opviera Inference
- Prompt compression
- CSV export
- RBAC
- AI Assistant
Pro
For engineering orgs that need analytics and governance.
- Extra requests
- $17 / 100K
- Max requests
- 50M / mo
- Projects
- Up to 25
- Data retention
- 90 days
Everything in Starter, plus
- Analytics dashboards
- Audit logs
- Custom roles
- Continuous abuse detection
Business
For scaled teams: unlimited projects, a year of history.
- Extra requests
- $15 / 100K
- Max requests
- No cap
- Projects
- Unlimited
- Data retention
- 365 days
Everything in Pro, plus
- Policy-level Direct Inference
- Priority support
- Unlimited projects
- 365-day retention
Enterprise
Run Opviera on your own infrastructure. Your cloud, your data, custom pricing and terms.
- ✓ Everything in Business
- ✓ Self-hosting in your cloud
- ✓ Unlimited requests
- ✓ Custom data retention
- ✓ Your own inference, no markup
- ✓ Dedicated support
The annual discount applies to the plan fee only. Included requests and extra-request pricing are the same as monthly. Estimate your cost →
Keep the AI savings. Cut the waste.
Illustrative model. A 200-developer org at a mid-range $800 / dev / month of uncontrolled AI spend.
Compression is on by default: the model assumes it takes ~13% off the cost, and that share grows the longer a session runs: the deeper the context, the more repeated history there is to compress. Token counts fall much further than cost, because the parts that dominate the cost (output tokens and cache reads) are not compressible.
Model routing sends the work that does not need a frontier model to an open one like GLM 5.2 or DeepSeek V4, so its saving comes off the residual cost rather than the starting one. Run it on today's models alone and the same org still lands at $1.57M: $350K saved, 18% lower.
Every plan, side by side
| Feature |
Free
$0/mo
Start free
|
Starter
$124.17/mo
Get started
|
Pro Popular
$1,665.83/mo
Get started
|
Business
$3,165.83/mo
Get started
|
Enterprise
Custom
Contact us
|
|---|---|---|---|---|---|
| Pricing & limits | |||||
| Plan fee | $0 | $1,490/yr | $19,990/yr | $37,990/yr | Contact us |
| Requests included / mo | 10K | 250K | 4M | 10M | Unlimited |
| Extra requests | Hard stop | $19 per 100K | $17 per 100K | $15 per 100K | Contact us |
| Max requests / mo | 10K | 6M | 50M | No cap | Unlimited |
| Projects | Up to 3 | Up to 10 | Up to 25 | Unlimited | Unlimited |
| Inference | BYOK only | BYOK or Opviera Inference | BYOK or Opviera Inference | BYOK or Opviera Inference | Your own infrastructure |
| Data retention | 3 days | 30 days | 90 days | 365 days | Custom |
| Controls | |||||
| Per-user keys and policies | |||||
| Project budgets | |||||
| Request log | |||||
| Smart model routing | |||||
| Opviera Harness (desktop app) | |||||
| Prompt compression | |||||
| CSV export | |||||
| RBAC | |||||
| AI Assistant | |||||
| Insight & governance | |||||
| Analytics | |||||
| Audit logs | |||||
| Custom roles | |||||
| Continuous abuse detection | |||||
| Policy-level Direct Inference | |||||
| Deployment & support | |||||
| Priority support | |||||
| Self-hosting | |||||
| Dedicated support | |||||
Estimate your monthly bill
Set your request volume and we'll pick the cheapest plan that fits. Using Opviera Inference? Turn it on, pick your models, and see the token cost at provider rates, with no markup.
Every model call through the gateway counts as one request.
Pay for tokens at provider rates, with no markup. Leave this off if you bring your own keys (BYOK).
What's typical?
Coding agents like Claude Code, Codex and OpenCode resend the conversation on every call, so a request carries about 15K–60K input tokens, and 70–90% of them are served from the prompt cache. Output is usually 300–1,500 tokens per call. The defaults here match a typical agentic coding session.
Extra requests are billed in 100K blocks, rounded up, at the end of each month. Annual plans pay the plan fee upfront, and extra requests and inference are still billed monthly. Inference estimates use each model's published per-token rates (global inference profile, where available). Cached input is priced at the cache-read rate, and cache-write premiums aren't included. Your invoice reflects actual usage.
Pricing FAQ
What counts as a request?+
A request is one model call that passes through the Opviera gateway, for example one turn of Claude Code, Codex or OpenCode talking to the model. Token count doesn't matter: a short call and a long one each count as one request. Failed requests are never counted.
What happens when I go past my included requests?+
On paid plans, traffic keeps flowing and extra requests are billed in 100K blocks: $19 per 100K on Starter, $17 on Pro and $15 on Business. Each plan has a monthly maximum (6M on Starter, 50M on Pro, no cap on Business), so a runaway script can't run up an unlimited bill. Usage charges (extra requests, Opviera Inference, AI Assistant) are invoiced monthly, or as soon as they reach $150 within the month. The Free plan hard-stops at 10K requests a month, and nothing is ever charged.
What's the difference between BYOK and Opviera Inference?+
With BYOK (bring your own key) you connect your own provider credential, such as a Claude subscription token or an Anthropic, OpenAI or OpenRouter API key. You pay the provider directly, and Opviera only charges your plan fee. With Opviera Inference (Starter and above) you don't need any provider account: we serve Claude, GPT, Grok, DeepSeek, Kimi, GLM, Qwen and more, and bill the tokens at the provider's published rates with no markup. Either way you keep the same budgets, analytics and abuse monitoring. For OpenRouter, Opviera sends "data_collection": "deny" with every request, so providers that may train on your inputs are skipped.
How does annual billing work?+
Annual plans pay the plan fee once a year upfront and get two months free, about 17% off: Starter is $1,490/yr, Pro $19,990/yr and Business $37,990/yr. The discount applies to the plan fee only. Included requests and extra-request pricing are the same as on monthly plans, and extra requests and any Opviera Inference usage are billed monthly.
What happens if I upgrade mid-cycle?+
Upgrades take effect immediately and are prorated, so you only pay the difference, and your renewal date doesn't change. We credit the unused part of your current plan fee and charge the prorated fee of the new plan for the rest of the cycle. Your new included requests apply right away.
What happens if I downgrade?+
You keep your current plan's features and included requests until the end of the cycle you've already paid for. Then your account switches to the cheaper plan automatically, and nothing extra is charged when you schedule it. If you change your mind before the switch date, the pending downgrade is simply cancelled at no cost.
How do hard-stop budgets prevent surprise costs?+
Budgets are checked before each request runs. When a user or project hits its daily or weekly token, session, or dollar cap, the request is blocked. Spend simply can't go over the limit you set.
Are failed requests or cache hits charged?+
Failed requests are never billed and never count toward your included requests. On Opviera Inference, prompt-cache savings are passed straight through to you at the provider's cache-read rate, never marked up.
How is Enterprise priced?+
Enterprise runs in your own infrastructure with unlimited requests and custom retention. You pay your AI provider directly for tokens and an agreed platform licence instead of per-request fees, and your data never leaves your environment. Talk to us →