Atlas applies rate limits to the requests that AI clients and agents send to the MongoDB Atlas Managed MCP Server. These limits help prevent a single user or agent from overwhelming the service, and help protect Atlas clusters and other downstream systems from unusually high request volumes. The limits apply regardless of which access model you use.
Important
MongoDB is still evaluating usage patterns for the MongoDB Atlas Managed MCP Server. The rate limits described on this page might change without prior notice.
Current Rate Limits
The current rate limits for the MongoDB Atlas Managed MCP Server are:
Limit | Rate |
|---|---|
Requests per IP address | 2,000 requests per minute |
Requests per user | 500 requests per minute |
Both limits apply to requests sent to the service. The per-IP limit applies to requests from a single IP address, and the per-user limit applies to requests from a single user.
Important
These rate limits apply only to requests sent to the MongoDB Atlas Managed MCP Server. The Atlas Administration API and the MCP Configuration API have their own rate limits, which apply separately. To learn about Atlas Administration API rate limits, see Atlas Administration API Rate Limits.
Service Account Access Token Limits
Atlas limits the number of active access tokens that each service account can hold. This limit is separate from the request rate limits described on this page. This limit applies to Programmatic Access Through MCP Configurations because Atlas creates a pair of Service Accounts for each MCP configuration.
Each service account supports a maximum of 100 active access tokens, and each token is valid for one hour. For details, see MongoDB Atlas Service Account Limits.
If you exceed the token limit, Atlas sends an HTTP 403 response with OAuth error code access_denied. The error_description contains the following error message:
Access Token limit reached. Please use the existing Access Tokens or create a new Service Account.
An agent that requests a new access token for each process, host, or session can reach this limit even when it stays within the request rate limits. Because tokens remain active until they expire or you revoke them, an agent that mints tokens faster than they expire cannot get a new token until earlier tokens expire.
To avoid hitting the token limit:
Cache each access token and reuse it until it expires, rather than requesting a new token for each request or each process.
Revoke tokens that you no longer need. For details, see Revoke Service Account Token.
Distribute agents across multiple MCP configurations so they do not share a single service account.
What Happens When You Exceed a Rate Limit
When you exceed a rate limit, the MongoDB Atlas Managed MCP Server returns an HTTP 429 (Too Many Requests) response and does not process the request.
If you receive a 429 response:
Wait before retrying. An immediate retry is likely to fail while you are still over the rate limit.
Retry with exponential backoff, gradually increasing the wait time before each retry.
Add a random delay (jitter) to avoid synchronized retries from multiple clients.
If the response includes a
Retry-Afterheader, wait at least as long as the header specifies.
Atlas does not currently provide a way to inspect the rate limits of the MongoDB Atlas Managed MCP Server or your current usage against those limits. You cannot request changes to these rate limits at this time. If the current limits are too restrictive for your use case, contact your MongoDB account team or MongoDB Support to flag that the limits do not meet your needs.
Learn More
To learn about Atlas Administration API rate limits and retry strategies, see Atlas Administration API Rate Limits.
To learn how the MongoDB Atlas Managed MCP Server secures access to your data, see Security, Governance, and Auditability.