Product
The AI control plane
One layer between your app and every model — find every LLM call, benchmark it on your real prompts, route it through a single gateway, and govern spend and access. Without rewriting your app.
Four pillars
Everything the control plane does
Each pillar is its own page — start anywhere.
Map
Your AI, mapped automatically
An IDE agent scans your codebase and registers every LLM call site as a workflow - file, line, model, and prompt. No manual instrumentation, no SDK wrapping.
ExploreGateway
One gateway for every model
Route all of it through one gateway with server-side keys and failover. Publish a new model or prompt from the dashboard, live on the next request, no redeploy.
ExploreEvals
Pick models with proof, not vibes
Benchmark candidate models and prompts against each workflow's real tests, scored by an LLM judge with six class-aware rubrics. Get a routing recommendation and catch quality drift before it ships.
ExploreGovern
Stay in control as you scale
Budgets that stop traffic at the cap, guardrails on inputs and outputs, team roles, and an audit trail - enforced at the gateway without touching your application code.
ExploreHow it works
From repo to routing in four steps
Connect
cran login then cran mcp install - one account login, then "use cran" in any repo.
Map
Scan registers every LLM call site. Live traffic merges on the same view.
Benchmark
Run your real prompts across every major model. Scored routing recommendations.
Operate
Publish model + prompt from dashboard. Proxied calls pick it up live.
See the whole stack.
Connect your repo and watch Cran map, benchmark, route, and govern every AI call.