# Decision Kernel

> The judgment API — a decision that was too expensive per row now fits in your pipeline.

Decision Kernel exposes TypeSafe Jev (a System One decision model) as a fast,
calibrated, low-cost judgment API. Send text; get department (triage), urgency,
spam, tone, and whether it needs human review — all in one request, ~0.5s, with
explicit confidence.

Built on Cloudflare Workers. Judgment proposals only — no money movement, no
financial/medical/legal accountability (classification + routing suggestions only).

## For AI agents

- **What this is**: a judgment API. Send a message; get 5 typed judgments with
  probabilities.
- **What it can do**: `POST /classify` returns triage, urgency, spam, tone, risk.
  `GET /openapi.json` has the machine-readable schema. `GET /agents.txt` has the
  capability card.
- **How to use it**: curl the endpoint or add it as an MCP/tool to your agent.
  No API key (starter) — see pricing tiers.

## Key files

- [openapi.json](openapi.json) — machine-readable API spec
- [agents.txt](agents.txt) — capability announcement
- [demo](/demo) — interactive web demo (free)

## Endpoints

- `GET /` — web demo
- `POST /classify` — the judgment API (5 judgments, parallel)
- `GET /health` — status
- `GET /openapi.json` — spec
- `GET /llms.txt` — this file
- `GET /agents.txt` — capability card
