The Logs tab in tool builder shows the invocation history for tools called by AI agents, MCP servers, or third-party applications. Logs appear only when the invoking resource's log mode isn't set to No logging. Use these logs to monitor tool runs, trace requests, and troubleshoot errors without re-running the tool.
Before you begin
- During the test mode, tool execution progress is available only in the Run Console and is not captured in the Logs tab.
- Execution logging must be enabled at the account level. See Introduction to execution logs.
- The log level for tool invocations is inherited from the invoking AI agent or MCP. You cannot set a log level at the tool level.
- Execution logs are available in non-production environments for accounts with a multi-environment license, depending on your account permissions.
- To view execution logs for tools that run in flows or APIs, use the flow's run logs or the API's request history. See View flow run logs or View API request and execution logs in API builder.
Open the Logs tab
- Go to AI Studio > Tools or Build > Tools.
- Select the tool you want to review.
- In the Tool builder, select the Logs tab from the top navigation bar right after the Build tab.
The Logs tab displays the invocation list, a table of past tool executions with latest invocations at the top.
Search and filter invocations
Use the options above the logs invocation list to narrow down the list:
- Search: Enter a search term to locate specific invocations
- Date range: Select a preset time window (for example, Last 30 days), or set a custom range
Set a custom date range
- Select the date range control, then select Custom range.
- In the Set custom range window, select a start date and an end date on the calendars.
- To narrow down the search further, select Time select, then set a Start time and End time. Time select is only available after you've selected a date.
- Optionally, select a different Timezone for the start and end times.
- Select Apply.
Select Clear to remove the custom range and return to a preset window.
You can also filter directly from the column headers. Select the filter icon on Resource type, Resource name, or Status to filter the list by that column.
Read the invocation list
Each row represents one invocation. The table includes these columns:
| Column | Description |
| Time | When the invocation started |
| Resource type | What triggered the invocation, such as MCP, AI Agent or 3rd party application |
| Resource name | The name of the triggering resource. For standalone invocations, this shows the caller's IP address instead |
| Status | Whether the invocation succeeded or returned an error |
| Duration (ms) | How long the invocation took to complete, in milliseconds |
| Trace key | The trace key associated with the invocation, if one was configured. See Set a trace key template |
| Triggered by | The user associated with the invocation. For a standalone tool invocation or MCP invocation, this is the user who owns the token used to invoke the tool. For an AI agent invocation, it's the account owner, regardless of which user runs the Agent. |
| Open trace view | An icon that opens the invocation's trace view. A Debug badge appears next to the Logs icon for any invocation logged at the debug level. |
Review an invocation
Select the icon in the Open trace view column for a row to open the trace view for that invocation.
Trace view has two panes:
-
Left pane: Displays a step tree for the invocation. A Tool input boundary step, the tool's internal steps (branches, lookups, imports, and other configured logic, in order), and a Tool output boundary step. If the tool calls another tool, the nested tool's steps appear inline, up to five levels deep.
- Right pane: Select any step in the left pane to see its input and output here. Connector steps (HTTP, NetSuite) show request and response details, nested tool steps show the nested tool's own input and output, transformation, mapping, and hook steps show a general input/output view.
Step detail for connector steps
The right pane has two tabs:
- Request: General details (the request URL and method), Headers, and Body for the outgoing request.
- Response: General details (the status code and content type), Headers, and Body for the returned response.
Headers and Body each display as raw JSON, with a copy icon to copy the contents.
For a lookup or import step that processes multiple records, select the step to see a child list of the individual records it processed. You can filter that list by status, and page through it if there are more records that fit on one screen. Selecting a child record shows that record's own input and output in the right pane.
Log level and available detail
The detail available in each invocation log depends on the log level of the invoking context (the AI agent or MCP that called the tool):
| Log level | What it captures | When to use it |
| No logging | Nothing is logged. | You don't want to store any data |
| Basic | Request URL, HTTP method, status code, duration, remote IP, and trace key. | Routine monitoring |
| Detailed | Everything in Basic, plus request and response payloads and headers for each step. | Data validation and troubleshooting |
| Debug | Complete execution detail, including inputs and outputs for internal step components such as mappings and transformations, and full request and response payloads. Time-limited. | Short-term, in-depth investigation |
To change the log level, update the setting on the invoking AI agent or MCP. See Introduction to execution logs.
Set a trace key template
Add a Trace key template to a tool’s tool input to define how the trace key is generated for each invocation. This field is optional. If you leave it blank, no trace key is generated for standalone, AI Agent, or MCP server invocations.
The template uses the same handlebars syntax as other tool input fields. The value it resolves to for each invocation appears in the Trace key column of the invocation list, so you can search for a specific run by that value.
Delete logs
Deleting logs permanently removes invocation data for a date range.
To delete logs:
- Open the Logs tab for the tool.
- Select Delete logs.
- Select a date range.
- Select Delete logs.
Delete logs are only available to account owners, admins, and users with managed access. Users with monitor only access don't see this option.
Connector support
Execution logs for tool invocations currently support HTTP and NetSuite connector types. Support for additional connector types is planned for future releases.
Limitations
- Test run results appear in the run console and are not stored in the Logs tab
- Logs are not available for invocations that ran before this feature was enabled
- Logs are not captured for deleted tool resources
- Log level cannot be set at the tool level. It is always inherited from the root execution context