Work Index

The Work Index (0-100) measures agent productivity per time window. It answers: is the agent actually doing work, or just burning tokens?

Scoring

Score	Status	Meaning
61-100	Working	High tool call rate, efficient token usage
26-60	Planning	Moderate activity, reading/exploring
1-25	Spinning	High token usage but few tool calls
0	Idle	No activity in this window

Formula

Score = min(100,
  (min(tool_density, 5) / 5) × 50      # Tool density weight
  + (min(token_efficiency, 3) / 3) × 30  # Token efficiency weight
  + (has_subagents ? 20 : 0)              # Delegation bonus
)

Where:

• Tool density = tool_calls / max(llm_calls, 1) — how many tools per LLM round
• Token efficiency = tool_calls / max(tokens / 1000, 0.1) — tools per 1k tokens
• Delegation bonus = 20 points if subagents were spawned (delegation = productive)

Special case: if llm_calls > 0 but tool_calls = 0, score is capped at 15 (spinning).

Use cases

Detecting spinning agents

An agent with Work Index consistently below 25 is likely:

• Stuck in a loop
• Generating long responses without taking action
• Failing silently (tool calls erroring out)

Comparing agent strategies

Run the same task with different models or prompts, compare Work Index:

• Higher density = more efficient tool use
• Higher token efficiency = less wasted context

Cost optimization

Agents with low token efficiency are spending tokens without results. Investigate:

openclaw traces:workindex --date 2026-03-12

Look for Spinning phases and check what the agent was doing during those windows.