Skip to main content

​
Restrictions

Restrictions limit assistant access and support allowlist/blocklist logic:
  • Connectors: Restrict accessible connectors (e.g., allow only Slack and HubSpot).
  • Actions: Block specific operations (e.g., destructive actions like delete_channel).
  • Users: Restrict which end-user IDs can access the server (useful for public setups).
Note: When updating, the full restrictions object replaces the previous configuration.

​
Server Descriptions

  • description: Context for the server, shown in dashboards and lists. Helps you organize and identify your servers.

​
Token Settings

  • tokenExpiresInDays: Set token lifespan (1-365 days). After expiration, the token URL will no longer work.

​
Response Handling

MCP responses are sized intelligently:
  • Large datasets: Response includes needsRefinement with suggestions to filter, search, or paginate.
  • Medium datasets (25-50 items): First 25 items returned with originalCount and displayedCount.
  • Small datasets (less than 25 items): Full response returned.
This ensures assistants are not overwhelmed.

​
Pagination and Rate Limits

  • List endpoints support limit and offset.
  • Responses include: total, limit, offset, hasMore.
  • Default rate limit: 100 requests/min per API key/token (sliding window).
  • Exceeding the limit returns HTTP 429.
  • Self-hosted servers can override limits via environment variables.

​
Input Validation

  • Server names: 1-100 characters, unique per user.
  • Token lifespan: 1-365 days.
  • Max request body size: 10 MB.
  • Arrays: Max 1,000 items per request or response.
These limits enforce reliability and prevent abuse.
⌘I