URL: /guides/cloud-audits

---
title: "Cloud audits"
description: "Log in, spend credits pay-as-you-go, and run cloud audits from the CLI, dashboard, or MCP with a shared issue tracker and published reports"
icon: "cloud"
---

Cloud audits add what your machine can't do on its own: AI page analysis, technology detection, browser rendering for JavaScript-heavy sites, a per-website issue tracker, and shareable published reports. This guide gets you logged in, explains how credits work, and shows the three ways to run a cloud audit.

Everything here builds on the free [local audit](/guides/local-audits-cli). The cloud runs the same rules and adds the credit-backed ones on top.

## Log in

Sign in once so your audits can reach the cloud:

```bash
squirrel auth login
```

This opens your browser to sign in. What runs next depends on the audit's coverage mode. Once you are signed in (free or Pro), audits default to `surface` coverage, which runs the cloud rules and merges their results into the same report as the local rules, spending credits as they run. Only anonymous, logged-out runs default to `quick`. Want a fast, free scan while signed in? Pass `-C quick`. Free accounts get 500 credits a month to spend; Pro includes 3,000. See [`squirrel auth`](/cli/auth) for API-key auth and credential precedence, and [`squirrel audit`](/cli/audit#coverage-modes) for the coverage modes.

## Credits are pay-as-you-go

Cloud features are priced in credits, small integers rather than dollars. Nothing is charged up front: credits are spent while an audit runs, and only for the features that run. Local audits never cost credits.

Free accounts get **500 credits every month**, no card required; Pro includes 3,000. Check your balance and the per-feature pricing anytime:

```bash
squirrel credits
```

A typical 50-page audit with the default cloud rules costs around 60 to 110 credits. Full pricing, plans, and top-ups are in [Credits and pricing](/cloud/credits).

<Tip>
You stay in control of spend. A per-audit cap (`max_credits_per_audit`) truncates work deterministically at the ceiling, and estimates above `confirm_threshold` prompt before spending. See [Spend controls](/cloud/credits#spend-controls).
</Tip>

## Run a cloud audit, three ways

The audit is the same underneath. Pick whichever surface fits how you work.

<Tabs>
  <Tab title="CLI">
    Run an audit with the cloud rules included. Signed in, the default `surface` coverage already runs them (and, when relevant, pages render in a cloud browser); the flag below just makes it explicit:

    ```bash
    squirrel audit https://example.com -C surface
    ```

    The start banner shows the estimated spend against your balance. See [`squirrel audit`](/cli/audit).
  </Tab>
  <Tab title="Dashboard">
    Trigger a run from [app.squirrelscan.com](https://app.squirrelscan.com): open a website and use its Run dialog. Cloud audits use the website's own [audit settings](/dashboard/audit-settings) (which rules run, external link checking, the spend cap), not your local `squirrel.toml`.
  </Tab>
  <Tab title="MCP">
    Your agent runs the audit as a tool call. Point any MCP client at the [hosted MCP server](/developers/mcp) and it calls `run_audit` with the URL, confirms the credit estimate, then polls `get_audit_status` and reads the report. The full agent workflow is in [Fix your website with an AI agent](/guides/fix-your-site-with-an-ai-agent).
  </Tab>
</Tabs>

## The per-website issue tracker

Cloud audits register the website and keep a tracker of its open issues, so findings persist across runs instead of scrolling past in your terminal. This is the backbone of the fix loop: an issue stays open until a later audit finds its rule no longer firing, then it closes on its own.

You work the tracker in the [dashboard](/dashboard), or your agent works it over MCP with `list_issues`, `get_issue`, and `comment_on_issue`. Recording a decision on an issue (why you are not fixing something, or what a fix changed) keeps the context with the finding for your team and the next audit.

## Published reports

When you are signed in and online, every audit auto-publishes to your dashboard as an **unlisted** report, giving you a shareable link plus issue history with no extra flags. Unlisted and private reports are free; only `public` reports cost 2 credits.

```bash
# signed in: auto-publishes (unlisted, free) and prints the URL
squirrel audit https://example.com

# make it public and listed (2 credits)
squirrel audit https://example.com --visibility public
```

Opt out of a single run with `--no-publish`, or turn it off for the project with `[cloud] publish = false`. Published reports are available in HTML, JSON, Markdown, plain text, XML, and the LLM format by appending an extension to the report URL. See [Reports and scoring](/reports#publishing-reports) for the details.

## The dashboard

[app.squirrelscan.com](https://app.squirrelscan.com) is where your cloud work lives:

- Published reports with health scores, and visibility controls (public, unlisted, private).
- The issue tracker for each website.
- Per-website [audit settings](/dashboard/audit-settings): rule selection, external link checking, and the spend cap.
- Billing and credits.

See the [Dashboard](/dashboard) guide for a tour.

## Keep audits running on a schedule

Instead of remembering to re-audit, have the cloud do it. Scheduled audits re-run a site daily, weekly, or monthly and update its tracker automatically, so regressions surface without you lifting a finger. A scheduled run that would exceed the website's spend cap is skipped with a notification, never a surprise bill. See [Scheduled audits](/cloud/scheduled-audits).

## Related

- [Cloud overview](/cloud) - what the cloud adds on top of local audits
- [Credits and pricing](/cloud/credits) - costs, plans, and spend controls
- [Hosted MCP server](/developers/mcp) - run cloud audits from your agent
- [Browser rendering](/guides/browser-rendering) - audit JavaScript-heavy sites
- [Fix your website with an AI agent](/guides/fix-your-site-with-an-ai-agent) - the full agent fix loop