Rate limits

The MCP server allows 10 requests per second for each user in each workspace. Every request to /mcp counts, including tool calls, tool listing, and session setup.

The limit is shared across all of your MCP connections in a workspace. Claude and Cursor connected as the same user in the same workspace draw from one budget. Connections in different workspaces, or by different users, have separate budgets.

When you go over

Whippy responds with:

HTTP/2 429
Retry-After: 1

Wait at least the number of seconds in Retry-After, then retry. Most assistants retry on their own. If yours shows the error instead, ask it to try again or to run fewer steps at once.

Public API limits also apply

Each tool call runs through the Whippy public API, so the API's own limits for your organization also apply. A tool can return a rate-limit error from the API even when you are under 10 MCP requests per second, for example while a script using an API key is busy in the same organization.

Avoid hitting the limit

• Ask for summaries over a filtered set ("contacts tagged Spring Hiring") rather than one-by-one lookups.
• Prefer filter_contacts with WQL over many show_contact calls.
• Turn off parallel tool calls in your client if it runs many at once.