Agent Logs in Sparkforge: Trace AI Agent Interactions

Agent Logs gives you one place to review how your AI agents behave across supported Sparkforge experiences. It pulls execution activity, conversation context, and step-level details into a single workspace so you can investigate issues faster.

What Agent Logs is

Agent Logs is Sparkforge's centralized logging and tracing experience for AI agents. It is live for Agent Studio and Voice AI today, with support for other AI products coming.

It is built around three diagnostic levels:

• A high-level view of agent activity.
• A conversation and execution timeline for a selected interaction.
• A granular step view with technical detail for debugging.

What you get

• Centralized visibility instead of isolated testing views.
• Faster troubleshooting from broad activity to a specific step.
• Better debugging context by comparing the response with the execution timeline.
• Performance awareness via step-level execution timing.
• Confidence to inspect prompts, logic, variables, and tool behavior.

Level 1: Global Activity Overview

The Global Activity Overview is a broad view of recent agent activity. Use it to monitor at scale, narrow down a problem, or locate an interaction worth inspecting.

Common items in this view:

Filter by product, agent name, or contact, and sort to find relevant interactions faster.

Level 2: Conversational Context and Timeline

After picking an interaction, you see the conversation alongside an execution timeline. Compare the agent's response with the steps that produced it, and trace the flow from the original user message through the final reply.

Level 3: Granular Step Execution

Step-level detail is where troubleshooting becomes precise. Inspect a single step to check whether an issue came from logic, timing, data handling, or a tool call.

Step output can be viewed as raw JSON or a parsed, readable view. JSON can be copied for sharing.

Useful step-level details:

• Model used for the step.
• Latency and execution timing.
• Timestamps.
• Input and output for the step.
• Prompt details where applicable.
• Additional technical metadata.

How to use Agent Logs

1. Open AI Agents > Agent Logs in your Sparkforge account.
2. Review the activity list. Filter by Agent Name, AI Product, Channel, or Status.
3. Open the interaction to see the conversation and execution timeline.
4. Select a message or step to drill into execution detail.
5. Use what you find to update your agent, then test again.

Metrics tab

The Metrics tab summarizes Agent Logs data into a customizable dashboard. Add, remove, drag, and resize widgets, then save multiple layouts. Track core KPIs (conversations handled, AI messages, average messages per conversation), performance (response time, top agents), operational health (top actions, average execution latency), and trends (busiest hours, channel activity over time).

FAQ

What is the main purpose of Agent Logs?

Understanding how an AI agent handled an interaction by combining conversation context with execution details in one place.

Is Agent Logs the same as Message Execution Timeline in Agent Studio?

No. Agent Logs is centralized logging and tracing across products. Message Execution Timeline is part of Agent Studio's testing and debugging experience.