Limits - Skan

Free allowances

Catalog tools report usage in an entitlement object on every response:

{ "tier": "free_guest", "used": 1, "included": 3, "remaining": 2, "upgrade_required": false, "more_in_app": "…" }

Capability	Included
Scan (full metrics)	Every completed scan can be read repeatedly, within the rate limits below
`recommend_products`	1 match per completed scan
`find_product`	3 lookups per completed scan; a scan is required. A lookup that finds nothing is not counted.
`analyze_ingredients`	10 per scan, or 5 per day with no scan

When an allowance is spent, the tool returns an empty result with upgrade_required: true and more_in_app guidance pointing to the Skan app. Starting another scan to reset an allowance is not permitted, and the guidance tells the model so.

Rate limits

Exceeding a limit returns rate_limited (or busy for concurrency) with retry_after_seconds. Service-wide ceilings return scan_limit_reached, session_limit_reached, capacity_paused, service_paused or admissions_paused.

Limit	Value
MCP requests per network address	60 per minute
Direct-photo scans per caller	60 per hour
Product calls per caller	10 per minute, 50 per day
Ingredient calls per caller	15 per minute, 40 per day
Deletions per network address	30 per minute (a separate budget, so deletion keeps working under load)
Account endpoint, per account	120 requests per minute, 10 scans per day
Admitted scans, service-wide	100 per day

Guest limits are keyed on a hashed network address, not on a person. An assistant platform calling from shared servers shares these budgets across its users. Values are configuration and may change.