Agent Time Tracking

Helm ADE automatically tracks time spent by AI agents during their sessions. This gives you visibility into how much automated work is being done on your projects without any manual tracking required.

How It Works

When an AI agent session runs, Helm ADE automatically:

Records time slices - Each segment of agent work is captured with start and end times

Links to context - Time is associated with the specific Project, Component, PRD, or Task being worked on

Stores metadata - The AI model used, session ID, and initiating user are all recorded

Viewing Agent Time

In Time Entries

Agent time entries appear alongside manual time entries with a purple "Agent" badge. Each entry shows:

• Duration of the time slice
• The entity it was linked to (Project, Component, PRD, or Task)
• The AI model used (e.g., GPT-4, Claude)
• The session ID
• When the work occurred

In Time Summaries

When viewing time summaries, agent time is shown separately from manual time:

• Blue bar - Manual/human time
• Purple bar - Agent time

This breakdown helps you understand the balance between human and automated effort on your projects.

In Time Rollups

Each entity (Project, Component, PRD, Task) shows a time rollup that includes:

• Total time across all sources
• Breakdown of human vs. agent time
• Number of entries from each source

Agent Time Metadata

Each agent time entry captures:

Field	Description
modelId	The AI model used (e.g., gpt-4, claude-3)
sessionId	The agent session that generated this time
initiatedByUserId	The user who started the agent session
entityType	What the time is linked to (project/component/prd/task)
entityId	The specific entity ID

Permissions

Agent time entries are system-generated and have special permission rules:

Action	Owner	Admin	Member	Viewer
View agent time	✅	✅	✅	✅
Edit agent time	✅	✅	❌	❌
Delete agent time	✅	✅	❌	❌

Note: Only administrators and owners can modify agent time entries. This ensures the accuracy of automated time tracking data.

AI Agent Tools

AI agents can query their own time data using the get_agent_time_summary tool. This provides:

• Total time for the current session
• Breakdown by entity
• Number of time entries

This is useful for agents that need to report on work completed during a session.

Understanding the Data

Session Splits

A single agent session may generate multiple time entries if it works on different entities. For example, an agent working on a PRD might:

Spend 10 minutes analyzing requirements

Spend 45 minutes implementing a task

Spend 15 minutes writing tests

Each of these would be captured as separate time entries linked to the appropriate entity.

Time Accuracy

Agent time is captured based on actual processing time, not wall-clock time. This means:

• Time waiting for user input is not counted
• Paused sessions don't accumulate time
• The duration reflects actual agent work

Related

• Manual Time Tracking - How to track your own time
• Viewing Time Rollups - Understanding time data at each hierarchy level