URL: /configuration/cloud

---
title: "Cloud Settings"
description: "Configure cloud features, spend limits, and browser rendering"
---

The `[cloud]` section of `squirrel.toml` controls [cloud features](/cloud): whether they run, how much an audit may spend, and whether pages are rendered with a cloud browser.

All cloud settings are no-ops when you're not logged in - cloud rules simply report `skipped`.

## Options

```toml squirrel.toml
[cloud]
enabled = true                # master switch for all cloud features
max_credits_per_audit = 1000  # hard spend cap per audit; 0 = unlimited
confirm_threshold = 50        # estimates above this prompt for confirmation
batch_size = 20               # pages per cloud service request
editor_summary = true         # Pro: auto-generate an editor's summary (8 credits)
# rendering                   # omit = auto (render when logged in); set "http" or "browser" to force
render_concurrency = 6        # concurrent browser-render jobs (1-8)
publish = true                # auto-publish to your dashboard when signed in
visibility = "unlisted"       # default visibility for published reports
```

### `enabled`

**Default:** `true`

Master switch. Set to `false` to disable all cloud calls for this project - cloud rules report `skipped`, publishing and rendering are unaffected by other settings.

### `max_credits_per_audit`

**Default:** `1000`

Hard cap on credits a single audit may spend. When the estimate exceeds the cap, cloud work is truncated deterministically and the rules that didn't run report `skipped` with reason `credit-cap-reached`. Set to `0` for unlimited.

### `confirm_threshold`

**Default:** `50`

In an interactive terminal, an estimated spend above this value prompts for confirmation before the cloud phase runs. Set to `0` to always confirm. Non-interactive runs (CI, agents, pipes) never prompt - they proceed bounded by `max_credits_per_audit`. Pass `--yes` to skip the prompt explicitly.

### `batch_size`

**Default:** `20` (also the maximum)

Pages per request to per-page cloud services. Lower it only if you hit request-size limits with very heavy pages.

### `editor_summary`

**Default:** `true`

Auto-generate an [editor's summary](/cloud/editor-summary) - an exec-email-style narrative plus point-form big-ticket items - at the top of the report. **Pro only**: free-plan audits skip it. Costs a flat 8 credits per audit; set to `false` to disable it for this project.

### `render`

**Default:** unset (auto)

Render strategy — the canonical control (supersedes `rendering`). Leave it unset and the strategy is decided per coverage: **quick** crawls HTTP-first and renders only what needs it; **surface/full** render every page (logged-in only, after a one-time consent prompt). Set it explicitly to pin behavior:

- `"off"` - plain HTTP only; never render, never spend render credits
- `"auto"` - **HTTP-first**: render *only* pages detected as client-side-rendered shells; static / server-rendered pages cost zero render credits
- `"all"` - render every HTML page at **2 credits per page**

`robots.txt`, XML sitemaps, and non-HTML assets are never rendered at any level — `"all"` means every crawled **HTML page**.

```toml squirrel.toml
[cloud]
render = "auto"   # only render the pages that actually need a browser
```

The `--render-mode off|auto|all` flag on `squirrel audit` overrides this for a single run (`--render` / `--http` are shorthands for `all` / `off`). See the [browser rendering guide](/guides/browser-rendering).

### `rendering` (deprecated)

**Default:** unset (auto)

Legacy fetch-mode toggle, kept for back-compat. Maps onto `render`: `"http"` → `off`, `"browser"` → `all`. Prefer `render`; if both are set, `render` wins.

### `render_concurrency`

**Default:** `6` (max `8`)

How many cloud browser-render jobs run at once during a rendered crawl. Render jobs run in squirrelscan's cloud, so the usual per-host politeness delay doesn't apply to job submission - but the render workers do fetch your site, hence the cap. Raise it to speed up large rendered audits; lower it if your origin is sensitive to parallel traffic.

Your plan also caps this value: **Free runs 1 render at a time, Pro runs 5**. The CLI checks your plan at the start of each rendered crawl and uses the lower of the two - a config value above your plan limit isn't an error, it just gets clamped.

### `publish`

**Default:** `true`

Auto-publish the audit report to your [dashboard](/dashboard) when you're **signed in and online** - independent of cloud rendering or enrichment. **Unlisted/private reports are free; only `public` costs 2 credits**, so the default (`unlisted`) auto-publish is free. Logged-out and `--offline` runs never publish. Set to `false` to never auto-publish for this project; you can still publish a single run with `--publish` (it overrides this). Skip publishing one run with `--no-publish`.

### `visibility`

**Default:** `"unlisted"`

Default visibility for published reports: `"public"` (listed and searchable, **2 credits**), `"unlisted"` (accessible via direct link only, free), or `"private"` (only you, when logged in, free). Override per-run with `--visibility`. The public charge applies once per report - flipping an existing report to public later charges the same 2 credits, then never again.

## Related

- [Cloud overview](/cloud) - what cloud features add
- [Credits & Pricing](/cloud/credits) - pricing and spend controls in depth
- [Cloud rules](/cloud/rules) - which rules use cloud services