For AI agents: The documentation index is at https://docs.digitalocean.com/llms.txt. Markdown versions of pages use the same URL with index.html.md in place of the HTML page (for example, append index.html.md to the directory path instead of opening the HTML document).

Action Gateway limits request rates and execution resources to manage capacity. Limits depend on the operation and the limits configured for your team. To discuss capacity for your workload or request a limit increase, contact support.

Rate Limits

• Tool calls are subject to requests-per-minute limits. Web search, web fetch, DigitalOcean tools, and third-party tools have separate rate-limit categories.
• Actor-specific limits can further restrict a workload. They do not increase the team’s limits.
• External providers can impose their own rate and concurrency limits in addition to Action Gateway limits.
• Each tool call in a parallel request or Python workflow is subject to the applicable limits. Grouping calls does not bypass them.

When a request is rate limited, follow its retry guidance and reduce the request rate or concurrency. Do not repeatedly resubmit a request without waiting. See Reliable Execution.

View and Lower Actor Limits

To view an actor’s limits, open Action Gateway in the DigitalOcean Control Panel, select the Actors tab, and click the actor. Its Limits section shows requests-per-minute (RPM) limits for these categories:

The Limits table in the Control Panel shows each category’s limits in these columns:

• Effective: The limit currently enforced for the actor, subject to the team’s limits.
• Configured: The actor-specific limit you set. Team default means the actor has no override for that category.
• Status: Shows Inherited when the actor uses the team default, Custom cap when an actor-specific limit applies, or Capped by team when the configured value exceeds the team’s limit.

To lower a limit for the actor, enter a whole number below its current effective limit in Configured, then press Enter or click outside the field to save. Check Effective to confirm the applied limit. To restore the team default, clear the field and save the same way.

Actor-specific limits cannot raise the team’s limits. Use the values shown for your actor rather than assuming a fixed default.

Request and Execution Limits

• Tool Search limits the queries and results in a request. Invoke Tool limits the number of parallel calls in a request. Use the schemas advertised by the session’s MCP endpoint to check the accepted request shape.
• Tool execution has timeouts. Code execution also depends on sandbox capacity and resource quotas. Its execution timeout includes time spent calling other tools.
• Each code execution starts in a fresh sandbox. Files, variables, and installed packages do not persist between executions.

Feature Restrictions

• Session configuration is set at creation. Create a replacement session to change its actor, selected tools, policy, preloads, output views, or VPC attachment.
• A session’s VPC attachment applies to code execution. It does not route every provider call through the VPC.
• Custom providers require a reachable HTTPS MCP endpoint using Streamable HTTP.
• Inline approval requires a client that supports MCP elicitation. Other clients must handle the separate approval flow.
• Provider permissions still apply to connected accounts. Gateway policy cannot grant scopes the provider has not authorized.