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

# IT - Fluency Agent Hardware & Performance Requirements

> CPU, memory, disk, and network impact of the Fluency Agent on end-user devices, with measured figures for macOS and Windows.

This page covers the hardware footprint of the Fluency Agent: CPU, memory, disk, and network impact on end-user devices. For installation instructions, see [IT - Deploying Fluency Agent](/it-deploying-fluency-agent).

***

## Summary

The Fluency Agent is a lightweight background application. Measured impact on both macOS and Windows is well within the capacity of any business laptop from the last several years.

**Recommended minimum device specification:**

| Component | Minimum                                         |
| :-------- | :---------------------------------------------- |
| RAM       | 4GB                                             |
| CPU       | Dual-core                                       |
| OS        | macOS 14 (Sonoma) or later, Windows 10 or later |

This is a conservative floor. Actual measured usage sits well below it, with comfortable headroom for other applications running at the same time.

***

## CPU & Memory

### macOS

| State                           | RAM       | CPU                                                     |
| :------------------------------ | :-------- | :------------------------------------------------------ |
| Idle (signed in, not capturing) | \~35–50MB | Near 0%                                                 |
| Actively capturing              | \~35–50MB | One CPU core, briefly (well under 1 second per capture) |

### Windows

| State                           | RAM         | CPU                              |
| :------------------------------ | :---------- | :------------------------------- |
| Idle (signed in, not capturing) | \~150MB     | \~10% of one core, on average    |
| Actively capturing              | \~150–185MB | \~15–20% of one core, on average |

Windows memory and CPU usage is somewhat higher than macOS. This is expected: the Windows build runs on Microsoft's WebView2 runtime (two OS processes: the agent itself plus a WebView2 helper process), while the macOS build uses the operating system's built-in WebKit framework, which carries less overhead per app.

In both cases, CPU usage returns to idle immediately after each capture. There is no sustained background load.

Windows figures were measured on a 13th Gen Intel Core i7 laptop with 32GB RAM (Windows 11, version 25H2). CPU percentages may run somewhat higher on older or lower-power hardware; memory usage is expected to be similar regardless of the host machine's specifications.

### Screen redaction

If your organization configures an application blacklist (to exclude sensitive apps like password managers or HR software from capture), the Agent redacts those regions from screenshots on the device before upload. This adds an additional, measured cost of about **1 millisecond per redacted region**, which is effectively unnoticeable and far smaller than the cost of processing the screenshot itself.

***

## Disk

| Item                     | Requirement                                                   |
| :----------------------- | :------------------------------------------------------------ |
| Application install size | \~15MB                                                        |
| Local logs               | Rotated daily, 3 days retained (typically well under 100MB)   |
| Offline queue            | Up to 1GB, only used if the device loses network connectivity |
| Screenshots              | Never written to disk, held in memory only until upload       |

**Worst case:** if a device is offline, events queue locally up to a 1GB cap; anything queued for more than 48 hours is discarded rather than accumulating indefinitely, and the entire queue uploads automatically once connectivity returns. **Typical case:** well under 100MB total, since the queue is normally empty and only logs and the application itself occupy disk space.

***

## Network Bandwidth

The Agent captures events based on user activity (each click), not on a fixed timer, so there is no constant background stream. Each event is a compressed screenshot plus a small amount of metadata, typically 300–800KB.

Daily upload volume scales with how actively a user works:

| Usage level | Approximate events/day | Approximate upload/day |
| :---------- | :--------------------- | :--------------------- |
| Light       | Under 1,000            | Under 0.5GB            |
| Typical     | \~2,000–4,000          | \~1–2GB                |
| Heavy       | 8,000+                 | \~4–6GB                |

Traffic is standard HTTPS and can be managed like any other cloud application in your network policy.

***

## Multi-Monitor & Display

Multi-monitor setups are fully supported with no additional configuration. The Agent automatically captures whichever display is active at the time of each click.

No dedicated graphics hardware is required. Screenshot capture uses the operating system's native APIs (ScreenCaptureKit on macOS, GDI on Windows).

***

## Support & Contacts

Oliver Farnill (CPO)

[oliver@usefluency.com](mailto:oliver@usefluency.com)

We will get back to you as soon as possible.