DEVELOPER GUIDE
Usage and limits
Read your current entitlements, credit balance, and operation costs before you run work.
Account usage
GET /api/billing/current-plan returns current plan and credit information. GET /api/user/limits returns resource limits and usage.
GET /api/billing/catalog provides the current credit catalog. Use these responses instead of fixed prices in your integration.
Manage subscription changes and checkout in the Funkel AI app. Payment finishes on Stripe's hosted page.
Operation costs
Costs depend on the operation and connected provider. Read the API operation and MCP catalog before paid work.
MCP browser approval applies to paid searches and non-read operations. Approval does not bypass quotas or sender limits.
Campaign scheduling and provider limits still apply after a campaign or agent becomes active.
Request limits
Limits vary by route. Funkel AI does not publish one global requests-per-minute allowance.
Agent message and attachment routes share a 20-request-per-minute authenticated-user limit.
Operator outcome and memory routes share a separate 20-request-per-minute authenticated-user limit.
Wait after a 429 response. Follow Retry-After when present, and avoid automatic retries of uncertain writes.
Handle failed requests
Check the HTTP status and response body. REST and MCP use different error formats.
Fix invalid inputs before retrying. Refresh state after conflicts. Review permissions and limits after denied actions.
Use asynchronous status routes when an operation returns a job or task. Request acceptance does not mean work has completed.