Rules

Rules that constrain what an agent may say and do.

Rules

A rule is a named constraint on agent behaviour — the guardrails that decide when an agent may act on its own, and when it must defer.

Rules are how you encode policy without changing code. For the reasoning that evaluates them against a decision, see NSR verified decisions.

Shape

Field	Type	Notes
id	string	Unique identifier
type	string	The kind of rule
name	string	Human-readable name
description	string	What the rule enforces
created_at	string	Creation timestamp
updated_at	string	Last modification

Endpoints

Method	Path
GET	/v1/rules	List
GET	/v1/rules	Get
POST	/v1/rules	Create
PUT	/v1/rules	Update
DELETE	/v1/rules	Delete

Example

curl https://api.stateset.com/v1/rules \
  -H "Authorization: Bearer $STATESET_API_KEY"

curl -X POST https://api.stateset.com/v1/rules \
  -H "Authorization: Bearer $STATESET_API_KEY" \
  -H 'content-type: application/json' \
  -d '{
    "name": "Escalate refunds over $100",
    "type": "escalation",
    "description": "Any refund at or above $100 requires human approval before execution."
  }'

Write the rule so a human can audit it. The description is what appears in a review queue when the rule fires, so “Any refund at or above $100 requires human approval” is more useful than “refund guard”.

Where rules are enforced

Layer	Mechanism
ResponseCX	Rules attached to an agent
Workflow Studio	Review gate and High-Value Action threshold
NSR	permit / deny / review effects with cited proofs
Temporal engine	action_rules — blocked, require_review, max_auto_amount

Related

• Verified Decisions
• Policy Engine & Decision Gate