One API.
Every AI model.
Infrastructure for AI.
GROUNDROCK is the programmable edge between your application and every frontier model. Route, cache, observe, secure, and fail over — from a single endpoint, without vendor lock-in.
One gateway.
Every provider. Zero lock-in.
Every request flows through the GROUNDROCK edge — routed, cached, observed, and secured — before it ever touches a model provider. Swap providers with a single header.
Everything you'd build yourself.
Already built. Battle-tested. Priced per request.
Universal Gateway
A single OpenAI-compatible endpoint that speaks to every provider. Switch models with a header, not a rewrite.
Smart Routing
Route by cost, latency, quality, region, or custom policy. A/B test providers and models with zero code changes.
Observability
Every request captured with token counts, latency, cost, and full I/O — searchable, exportable, and streamable.
Semantic Cache
Embedding-based cache with per-project TTL. Cut spend by up to 68% on repetitive workloads.
Automatic Fallbacks
Multi-provider retries with typed error taxonomy. If OpenAI is degraded, we're already on Anthropic.
Governance & RBAC
Per-project keys, spend caps, model allowlists, and audit logs. SSO/SAML and SCIM on Team+.
Analytics
Real-time dashboards for spend, tokens, latency, error rate — sliced by project, model, or environment.
Guardrails
PII redaction, prompt-injection detection, and jailbreak filters — before the request leaves your edge.
Three steps.
From npm install to production.
Install
One SDK. OpenAI-compatible. Drop it in beside your existing client.
npm install @groundrock/sdkPoint at GROUNDROCK
Change one base URL. Keep your existing OpenAI code. Every provider, one call.
import { Groundrock } from "@groundrock/sdk";
const ai = new Groundrock({
apiKey: process.env.GROUNDROCK_KEY,
});Ship
Route, cache, observe, and fail over — automatically. Ship on day one.
const res = await ai.chat({
model: "openai/gpt-5",
fallback: ["anthropic/claude-opus-4"],
messages: [{ role: "user", content: "Hello." }],
});Model-agnostic infrastructure.
Priced like a utility.
GROUNDROCK runs on the same primitives that power modern cloud infrastructure — anycast routing, deterministic caching, first-class observability, region-aware failover. You get the reliability of AWS with the ergonomics of a single API.
Built for the enterprises
already running production AI.
Deploy inside your cloud. Retain your keys. Prove compliance with a cryptographically-signed audit trail. Your CISO's Friday will be quiet.
BYO Cloud
Deploy inside your VPC on AWS, GCP, Azure, or bare metal.
Dedicated Regions
Region-pinned inference with data-residency guarantees.
SSO & SAML
Okta, Entra, Google Workspace, and SCIM provisioning.
Private Networking
VPC peering, PrivateLink, and mTLS between your app and the edge.
Guardrails at the Edge
PII redaction, prompt-injection filters, jailbreak detection.
Signed Audit Logs
Every request cryptographically signed and streamed to your SIEM.
Pay per request. Scale to zero.
No seats. No hidden overage.
Everything you need to prototype against every provider.
- 10,000 requests / month
- All providers, all models
- Basic observability (24h retention)
- Community support
Production-ready limits and analytics for solo teams.
- 1M requests / month included
- $0.0002 per request over
- Analytics · 30d retention
- Automatic fallbacks
- Email support
For teams shipping AI features at scale.
- 25M requests / month included
- $0.00015 per request over
- Semantic cache included
- Analytics · 90d retention
- SSO / SAML
- 99.99% uptime SLA
- Priority support
BYO cloud, dedicated regions, and signed compliance.
- Unlimited volume
- BYO cloud (AWS · GCP · Azure)
- Dedicated inference regions
- SOC 2 / HIPAA / GDPR reports
- SCIM · RBAC · Audit logs
- 99.999% uptime SLA
- Named SRE + solutions architect
We're building the AI cloud.
Here's what's shipping next.
Answers, up front.
Start building on GROUNDROCK.
Every AI model, one endpoint, production-grade primitives. 10,000 requests free forever — no credit card required.