Agent Activity

Documentation Index

Fetch the complete documentation index at: https://docs.maadify.com/llms.txt

Use this file to discover all available pages before exploring further.

​

What you can monitor

The Agent activity page shows what happened when users chatted with agents or when workflows ran from triggers. Use it to review the path an agent took, the decisions it made, the tools it called, the files or sources it used, and any approvals or errors that occurred. You can see:

• Agent and sub-agent steps in the order they ran.
• Tool executions with inputs, outputs, and status.
• User messages that started or continued the workflow.
• Rules and conditions that controlled the next step.
• Approvals that are waiting, approved, or rejected.
• Sources, files, and browser recordings attached to a step.
• Errors with the failing step and related context.

​

Open the flow view

Select an activity record to open the run details. The Flow view shows the workflow path as a connected diagram.

​

Review the visual flow

The Flow view shows the run as a connected diagram. Each node represents a user message, parent agent, sub-agent, tool execution, or workflow end state. Nodes include:

• Type badges such as Agent, Sub-Agent, Tool, and User.
• Status badges such as Success, Error, Awaiting Approval, Approved, and Rejected.
• Expandable agent groups when a parent agent contains sub-agent or tool steps.
• Configuration links on agent nodes when the agent configuration is available.
• Approval end nodes that show whether a run is awaiting approval, approved, or rejected.

The flow also includes zoom controls, a minimap, and a fullscreen option so you can inspect larger workflows.

​

Understand rule paths

Lines between nodes can show the rule or condition that caused the next step to run. These labels help you understand why the workflow followed a specific path. Rule labels can reference:

• Last message content.
• Runtime inputs from the workflow.
• Tool content returned by a previous tool.
• Last agent or tool executed.
• Context values stored during the run.
• Skip conditions that prevented a step from running.

​

Filter, search, and sort activity

Use filters to narrow down activity and find the exact step or workflow run you need.

• Search by agent name, input, output, or company.
• Sort by date, agent name, status, or company.
• Filter by status to focus on approvals, errors, or successful steps.
• Date filters let you isolate activity over a specific window.

​

Inspect step details

Click a node to open the step detail panel. Use the panel to inspect what the step received, what it returned, and what supporting data was attached. Depending on the step, the detail panel can show:

• Input for the message, prompt, or tool parameters sent into the step.
• Output for the agent response, tool response, JSON result, markdown, or CSV.
• System for the system message used by a sub-agent.
• Context for usage and context details, such as token usage, message window size, memory context, and compression details.
• Elements for files, sources, and browser recordings attached to the step.

The panel can be expanded when you need more room to review long inputs, outputs, or source content.

​

Review Context

Use the Context tab to understand what the agent had available before it responded. This is useful when you need to explain why an agent answered a certain way, why context was removed, or whether memory changed the message window. The Context tab can show:

• Token usage for the model, including total and actual prompt and completion tokens.
• Context window stats such as pre-transform messages, post-transform messages, removed messages, compressed content, tool calls, tool responses, memory context, and message senders.
• Transform details that show how message transforms changed the context before the model ran.
• Pre-transform messages that show the message window before context processing.
• Post-transform messages that show the message window after context processing.

Transform details can include memory storage, compression, trimming, visibility filters, and memory injection. Use the pre-transform and post-transform message lists together to see exactly what changed before the next step ran.

​

Review elements, sources, and recordings

The Elements tab shows supporting artifacts created or used by a step. You may see:

• Files that can be downloaded when a workflow created or attached a file.
• Sources that show documents, records, or links used to support an answer.
• Session recordings when an agent acted in a browser and captured a replay.

Click a source badge to view its content. If the source is a URL, it opens in a new tab. Click a session recording to replay the browser activity.

​

Approve or reject actions

Approval steps appear in the run timeline and in the flow. If a workflow is waiting for approval, the detail panel shows a Review Approval action and the flow shows an approval end node.

​

Approval statuses

• Waiting: approval required before the workflow continues.
• Approved: the action was approved and the workflow resumed.
• Rejected: the action was rejected and the workflow stopped or rerouted.

​

Workflow runs and trigger executions

The Workflow runs view shows trigger executions with request and response payloads. You can filter, sort, and re-run selected executions.

​

Re-run selected executions

Select one or more workflow runs to re-run them. Use this for retrying transient failures or validating a fix.

​

Error visibility

When an error occurs, you can see:

• The step that failed.
• An Error badge on the node or list item.
• Input and output context at the failure point.
• Any related files, sources, or browser recording.
• Links to related notifications when available.

​

Troubleshooting