AI Agent Feature Invocation Timeout After 60 Seconds During Agent Capability Execution
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
a week ago
Hi Team,
We are encountering a timeout issue while invoking a capability from an agent.
Error Message:
Plain Text
Processing for feature invocation cc234d083ba4571476411fdc73e45a41 for capability f6f9e6e71bc38210f4200d81ec4bcb30 timed out after 60000 milliseconds.
Environment
Product: Now Assist / AI Agent Framework
Capability ID: f6f9e6e71bc38210f4200d81ec4bcb30
Feature Invocation ID: cc234d083ba4571476411fdc73e45a41
Issue Details
I am creating a flow using Agent by calling a tool.
The agent successfully starts processing the request.
The flow is getting generated but the response is not showing in the agent.
When i checked the logs it is showing timeout error
The underlying capability performs a long-running operation (flow generation).
After exactly 60,000 ms (60 seconds), the invocation times out.
The operation appears to continue in the background and may complete successfully, but the agent does not receive or display the final response.
We are trying to understand whether the timeout is configurable and if there is a supported approach for long-running capability executions.
Questions
Is the 60-second feature invocation timeout configurable?
Is there a system property or configuration for increasing this timeout?
Are there any recommended patterns for handling long-running operations in AI Agent capabilities?
Is this timeout enforced by the AIA ReAct Engine, Agent Framework, or capability execution layer?
Are there logs or diagnostics that can help identify the exact timeout source?
Troubleshooting Performed
Verified capability execution.
Confirmed the process requires more than 60 seconds for complex flow generation.
Checked agent logs and observed the timeout occurs consistently at 60 seconds.
Any guidance or documentation would be appreciated.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
a week ago
Your flow generation can work reliably, because there is a clear pattern for long-running work in agents. The documentation I found does not list a setting for this 60-second limit, so the dependable route is to design around it rather than raise it. A Community thread reports the same 60-second limit when a Skill Kit skill is called from Flow Designer. That user’s skill ran fine on its own, and the advice was to move to the latest patch.
To let the agent handle long jobs:
- Start flow generation in the background, for example from an asynchronous flow, a scheduled job or an event. Have the tool return straight away with a reference to a tracking record.
- Add a second tool that reads the status from that tracking record. Instruct the agent to check it and report the result when it is ready.
- Store the generated flow link or result on the tracking record, so the agent always has something to show.
- Where the requirement allows, split a complex request into smaller generation steps so each call finishes comfortably within the limit.
These tables help you diagnose each run:
- sn_aia_execution_plan shows each execution once you add the Gen AI usage log column, as a Community answer suggests.
- sys_gen_ai_usage_log shows the invocation and its usage entry. It may need adding to glide.ui.permitted_tables for reporting.
- sys_gen_ai_log_metadata gives success and error details, and needs the same property.
ServiceNow Support can confirm exactly where the timeout is enforced and whether your patch level affects it. Send them the feature invocation ID and capability ID. Keep your Now Assist plugins and patch level current, since other users report that helps, and build the background job in a sub-production instance first.
Sources: