> ## Documentation Index
> Fetch the complete documentation index at: https://docs.flowfield.sh/llms.txt
> Use this file to discover all available pages before exploring further.

# Metrics

> Track tokens, cost and running time across your workspace and tasks.

Open **Metrics** in the sidebar to compare activity across tasks, agent types, harnesses or models. Metrics supports Codex, Claude Code and Pi.

![Workspace Metrics with shared filters and totals](https://mintcdn.com/flowfield/ADLarR3F5Xq-qO1B/images/metrics.png?fit=max\&auto=format\&n=ADLarR3F5Xq-qO1B\&q=85\&s=c7616cf22497e404a7e0da46b2c67932)

## Explore workspace activity

Choose a **Period**, or enter **From** and **To** in your local time and select **Apply**. Use **Activity** to select all work, the Coordinator or a task, and **Group by** to compare agent types, harnesses or models.

The charts and task table share these filters. Click a day or hour to narrow the range; use arrow keys and Enter when navigating by keyboard. Sort the Tasks table by Tokens, Cost or Duration, then open a task for its details. Bookmarks retain your selection.

Tokens and Cost use stacked bars. Duration uses separate lines because parallel work overlaps and should not be added together.

![Token and cost bars for the selected period](https://mintcdn.com/flowfield/ADLarR3F5Xq-qO1B/images/metrics-cost.png?fit=max\&auto=format\&n=ADLarR3F5Xq-qO1B\&q=85\&s=4a7dafe005ce2cfca1a139e3baa3e2f9)

## Inspect a task

Tasks show lifetime **Tokens**, **Cost**, **Duration** and **Corrections**. Hover, focus or tap a callout for details and the Worker/Reviewer breakdown. On narrow screens, scroll the row to see all four.

![Task metrics and recorded work](https://mintcdn.com/flowfield/ADLarR3F5Xq-qO1B/images/task-metrics.png?fit=max\&auto=format\&n=ADLarR3F5Xq-qO1B\&q=85\&s=990b55899f6c7efebc40a6d8043f5854)

Corrections counts work started to address feedback, excluding retries and answer continuations.

## Understand the numbers

* **Tokens** are reported by the harness, including cached input without double-counting.
* **Cost** estimates reported input, output and cache usage at each model’s USD API rates, including helper models. Missing Codex cache detail uses the input rate. Tooltips show pricing dates; saved costs keep their rates. Values below one cent show `<$0.01`.
* **Duration** counts overlapping running time once and excludes idle gaps.

Cost totals include the usage that can be priced; their tooltip identifies incomplete coverage. Unavailable measurements show `—`.

## Investigate agent context

Ask the Coordinator to inspect an execution's context measurements when you want to understand large responses or repeated document reads. These measurements show the sizes of prompts, tools and outputs; [MCP reference](https://docs.flowfield.sh/mcp-reference.md#read-execution-measurements) explains how to retrieve and interpret them.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.