KYI Framework is Abstraxn’s layer for Know Your Intent — proving that an AI agent was allowed to act before it does something irreversible (transfer, order, MCP tool call, etc.). Warrant is the first module inside KYI Framework. It gives you:
  • Mandates — rules the owner signs (spend limits, allowed addresses, etc.)
  • check() — ALLOW or DENY before the action runs
  • Receipts — signed proof of every decision (for audits and compliance)

​
Who this is for

  • Teams building AI agents that move money or call sensitive APIs
  • Products that need a clear audit trail (“why was this transfer allowed?”)
  • Anyone who wants policy before execution, not only logs after the fact

​
The idea in one sentence

The owner sets rules once → the agent asks Warrant before each action → Warrant returns ALLOW or DENY plus a signed receipt.

​
How it fits with Agent Kit

Example: your agent wraps an MCP transfer tool. Call warrant.check() first. If DENY, never call the real transfer.

​
Two API keys (important)

When you create a mandate, Warrant returns a mandate API key once. Store it securely — runtime agents use that key, not your app key.

​
API base URL

All Warrant paths live under:

​
Choose your path

Dashboard setup

Enable KYI Framework in the dashboard, pick receipt chain, view mandates.

SDK quickstart

Install @abstraxn/warrant, create a mandate, call check().

React quickstart

MandateForm + deterministic Readback for mandate signing UI.

Quick start

Copy env vars and run your first check() in minutes.

How it works

Simple walkthrough — mandate → check → receipt → verify.

Verify receipts

Auditor path — prove a receipt is real without trusting Abstraxn alone.

​
Packages

Doc guides: SDK quickstart · React quickstart · Verify receipts

​
Next steps

  1. Quick start — env vars + first check
  2. Dashboard setup — turn on KYI for your app
  3. SDK quickstart — full mandate create flow
  4. React quickstart — mandate signing UI
  5. Agent Kit integration — gate MCP tools