What usage limits and spending caps does CMD+K enforce?
CMD+K has two kinds of limits: plan-level quotas (included monthly credits on widget plans, request buckets on the MCP gateway) and operator-set hard caps (the maximum monthly USD you authorise us to bill you). Hard caps trip the same suspension gate across CMD+K Layer, MCP Gateway, and Monetize: once you cross the cap, the next request returns a graceful blocked response instead of running. Caps reset at the start of each billing period.
Plan quotas
- CMD+K Free Trial — 14 days, 150 credits/month, 1 widget.
- CMD+K AI Adopter — 1,500 credits/month included; extra usage from your prepaid wallet.
- CMD+K AI Pilot — premium models unlocked, whitelabel, SSO, 6,000 credits/month included.
- MCP Free — 1 endpoint, 2,500 requests/month, no PAYG.
- MCP Paid — $9.99/endpoint/month with 10k included; PAYG $0.50/100k after.
Hard monthly caps
Every paid subscription has a hard_monthly_cap_usd. Stripe billing thresholds ($20) trigger an invoice early so we can verify the payment method is good; the operator-set cap is enforced server-side by the suspension gate. You can edit the cap inline on the billing page for both CMD+K and MCP.
Per-end-user caps and the tier claim
On top of the workspace cap, Monetize lets you cap each end-user's monthly spend. A visitor's effective cap is resolved in this order — the first match wins:
- Per-user override — a cap you set on that visitor in Monetize → Users.
- Tier — the
tierclaim in the visitor's widget JWT (your own plan label, e.g."free","pro"), matched against the per-tier caps you configured. - Default — the workspace-wide default cap.
tier is distinct from the lifecycle plan claim (trial | paid | free), which only steers the activation agent and never affects caps. See the JWT claim reference and Per-user spending caps.
Suspension semantics
- MCP Gateway — requests return HTTP 402 with a JSON body explaining the block.
- Widget — assistant turn ends with a 'limit reached' message; no model call billed.
- Monetize — per-end-user caps block that specific user; other users keep working.
- Past-due payment — workspace-wide suspension across all products until the invoice clears.
Next steps
FAQ
- What happens at the hard cap?
- Requests return HTTP 402 (MCP) or a 'limit reached' message (widget). No AI call is billed past the cap.
- Can I get an email before I hit the cap?
- Yes — workspace admins are emailed at 80% and again at 100% of the configured hard cap.
- Do caps cover Monetize forwarding too?
- Yes. Per-end-user caps inside Monetize use the same gate; see Monetize → Per-user spending caps.
- Can different customer plans get different per-user caps?
- Yes. Put your own plan label in the widget JWT's tier claim (e.g. "pro") and configure a cap per tier under Monetize → Caps. A visitor's cap resolves as per-user override → tier → workspace default. tier is not the lifecycle plan claim (trial | paid | free), which only steers the activation agent.