● Detecte/docs

Detecte

Detecte is runtime security for AI agents. You verify an agent before you deploy it. You verify every action it takes after.

It is a single API. Five lines of code in your agent. A dashboard your security team will actually open.

Why this exists

LLMs are non-deterministic. Agents have tools. Tools touch real systems — money, data, customers. The combination is new, the failure modes are new, and the existing security stack does not see them.

Detecte sits in the loop. Before any agent action runs, you ask Detecte: "is this allowed?" Detecte checks your policies, your behavioral baselines, your KYA verification, and answers in <100ms.

What you'll build today

  • Sign up at app.detecte.xyz and get a key in 60 seconds
  • Drop five lines into your agent — see decisions appear in the dashboard live
  • Write a policy that blocks transfers over $10,000
  • Wire a webhook for real-time alerts
  • (Later) Run the KYA test battery against your agent's prompts

Start with the Quickstart.

How it's structured

The product has four pieces you'll touch:

  • The SDK — @detecte/sdk for TypeScript, detecte for Python, @detecte/cli for the terminal.
  • The API — direct HTTP if you don't want an SDK.
  • The dashboard — agents, policies, decisions, incidents, audit log.
  • The webhooks — real-time events into your systems.