The decisions of record

Developers

Built for the agentic future.

Vrrdict is API-first by design — the same verdict engine the app runs on is callable programmatically. When your agent needs a defensible go/no-go, it can ask Vrrdict instead of hand-rolling a weather prompt.

API-first

Every capability, an API
There are no internal-only endpoints. The web app talks to the same authenticated API third parties and the MCP server will use — verdicts, rules, signals, audit, all of it.
Verdicts agents can read
A verdict comes back as a structured object — state, the triggering rule, reasoning, confidence — not prose to re-parse. Branch on it directly.

The verdict, two ways

Saturday game
Rain check · evaluated 14:02 UTC
Proceed
your rule · rain ½″
ThuFriSat

Your Rain check rule — 0.04 in by noon — under your ½ in line, nothing firing.

CONF 86% · sources: NWS

the card your users see ↓ the object your agent reads

Example — a verdict object
{
"state": "proceed",
"triggering_rule": "Rain check — Saturday game",
"reasoning": "0.04 in by noon, under your ½ in line.",
"confidence": 0.86,
"evaluated_at": "2026-05-24T14:02:11Z"
}

MCP server — coming

A tool contract, not a prompt

We are building an MCP server so an agent picks Vrrdict as a first-class tool — typed inputs, a verdict object out, the audit trail behind it.

Pricing for agents

Per-verdict, not per-seat

Agent traffic will be metered per call / per verdict, so a fleet of agents can lean on the engine without a seat for each. Details at launch.

Decide once. Your rules take it from here.

Building an agent that has to decide? Let's talk.

The API is in private development. Tell us what you're building and we'll bring you in early.

Developers & API — Vrrdict