Browse the docs
Developers
Rate limits and usage
How many requests the ASDK API and MCP server accept, how answers count toward your plan, and how to handle 429 responses and Retry-After.
Last updated
On this page
Limits keep the API fast for everyone. When you reach one, the API answers 429 with the problem type rate-limited (or answer-limit-reached for monthly answers) and a Retry-After header with the seconds to wait.
Request limits
| What | Limit |
|---|---|
REST API (/v1/…) | 300 requests per minute per client IP address |
| MCP server, per API key | 120 requests per minute |
MCP ask tool, per API key | 20 calls per minute |
| MCP server, per client IP address | 300 requests per minute, including requests with a wrong key |
| Webhook test events and resends | 60 per hour per workspace |
Answers
Every answer an agent writes counts toward the plan's monthly answers: the widget's answers, POST /v1/sites/{siteId}/ask and the MCP server's ask tool alike. Knowledge search (search_knowledge) is not an answer. When a workspace reaches its monthly answers, ASDK notifies its owners; past 110 % of them, server-side answers stop with 429 answer-limit-reached until the next month (UTC), and the widget tells visitors it can't answer more questions. The pricing page lists the answers of each plan.
Handle 429
- Wait for the seconds in
Retry-After, then retry the same request. - Spread scheduled jobs out instead of sending a burst at the top of the hour.
- Use webhooks instead of polling for new leads and conversations.
- Use one key per integration, so one busy integration does not slow down the others on the MCP server.