routeva

Simple, honest pricing

You bring your provider keys — we don't mark up the LLM cost. Pay for the gateway, not the tokens.

Free
$0/ forever
For individual developers and side projects.
10,000 requests / month
  • All 8 providers
  • Smart routing (weak / strong / automix)
  • 9 built-in guardrails
  • Semantic + exact cache
  • 1 project
  • Community support (GitHub issues)
Get started
Most popular
Pro
$29/ per month
For teams shipping to production.
500,000 requests / month
  • Everything in Free
  • Unlimited projects
  • Eval-driven smart routing (pluggable SmartScorer)
  • Continuous eval loop (online scoring + annotations)
  • Priority email support
  • Team seats (up to 5)
  • Custom retention policies
Start Pro trial
Enterprise
Custom
For regulated workloads and larger teams.
Unlimited
  • Everything in Pro
  • Private VPC deployment (self-host or dedicated)
  • SSO / SAML
  • SOC 2 documentation
  • SLA + 24/7 support
  • Custom guardrail hooks
  • Dedicated Slack channel
Contact us

Frequently asked

Are token costs included?
No — you bring your own provider API keys (Bedrock, OpenAI, etc.) and pay them directly. Our fee is for the gateway itself: routing, guardrails, cache, and the eval loop. This means you keep provider relationships and negotiated discounts.
What counts as a “request”?
One inbound call to /gateway/v1/chat/completions. Cache hits, guardrail-blocked calls, and internal retries count once. Streaming responses count once regardless of chunk count.
What happens when I hit my quota?
On Free: requests return HTTP 429 until the next month starts. On Pro: overage is billed at $0.20 per 1,000 requests up to a soft cap, then throttled. You can set a hard cap in Settings.
Can I self-host?
Yes — the OSS build is Apache 2.0 and lives at github.com/routevaai/routeva-gateway. All non-multi-tenant features are identical between OSS and the hosted plan. Enterprise self-host adds SSO + multi-tenant + support.
How is Pro different from a raw Bedrock/OpenAI account?
Multi-provider fan-out, 9 built-in guardrails, semantic cache with $ tracking, eval-driven routing that shifts traffic to weak models when your evals allow. On typical chat/agent workloads customers see 40–70% cost drop vs. always-strong.
Is there a trial?
Yes — 14 days of Pro with no card required. If you don't upgrade, you drop to Free automatically.

Not sure which tier fits?

Start on Free — most teams don't need Pro until they're past the pilot stage.

Get started for free →