Each client gets their own policy, budget and approved models. You get one screen across all of them, a statement you can invoice from, and a read-only login you can hand the client.
5,000 free executions · $1 model credit · 14 days · No card required
const client = new OpenAI({
apiKey: process.env.PLANVERITY_API_KEY,
baseURL: "https://api.planverity.ai/v1"
});Point an OpenAI-compatible client — or an n8n AI node — at Planverity and nothing else changes. Model access, routing, budgets and execution records all happen behind the one endpoint.
Any OpenAI-compatible client works unmodified, including streaming and tool calls — verified against n8n's MCP integration.
One organisation per client, each with its own policy, budget and approved models. Isolation is enforced in the query, so a key from one client gets a 404 on another client's execution — not a 403, which would confirm the record exists.
Explore multi-client →Planverity selects the lowest-cost model that meets your quality, budget and policy rules.
Explore routing →Bring your own keys, add prepaid model credit or connect local infrastructure.
Explore model access →Atomic reservations stop simultaneous requests from spending beyond your limit.
Explore cost controls →Continue through approved providers when a model times out, rate-limits or becomes unavailable.
Explore reliability →Models see only the tools permitted and relevant to the current request.
Explore MCP governance →See what ran, why it was selected, who funded it and what it cost.
Explore execution records →Including where it loses: Planverity is Pareto-efficient but does not dominate a single good mid-tier model on general traffic. View the research →
one organisation
per month · 3 organisations
per month · 25 organisations
Start with 5,000 free executions and $1 in model credit. No card required.