Pre-Send Email HTML Compatibility Checker

Check email HTML before you send it. The Actor finds design breaks in Outlook and Gmail, and in ten other clients, without a browser and without a render farm. Send it your email HTML. Get a client-by-client score, a list of findings, and a send verdict.

What it does

For each email, the Actor:

1. Reads the HTML you supply, either inline or from a public URL.
2. Tests the HTML and CSS against a static rule matrix for up to 12 email client profiles (Outlook desktop, Outlook.com, Gmail web/Android/iOS, Apple Mail macOS/iOS, Yahoo, AOL, Samsung Mail, Thunderbird, Orange.fr).
3. Sends one bounded HEAD request for each unique image and link that the HTML names, to check that the asset is reachable, served over HTTPS, and not oversize.
4. Reports structure flags: missing alt text, a missing preheader, table layout use, media query count, and the Gmail 102 KB clip limit.
5. Scores each client from 0 to 100 and gives one verdict: SEND_OK, SEND_WITH_RISK, or DO_NOT_SEND.

The Actor never renders the email and never opens a browser. It reads the supplied HTML as text and sends network requests only to asset URLs that the HTML itself names.

Input

See .actor/input_schema.json for the full schema. The main fields:

Field	Type	Default	Purpose
emails	array	the example email	Each item needs name and one of html or htmlUrl. Maximum 200 items, 2 MB of HTML each. If you leave this field empty, the Actor checks the example email from the default input.
clientProfiles	array of string	all 12 profiles	Which client profiles to score.
checkRemoteAssets	boolean	true	Send HEAD requests to check images and links.
maxAssetsPerEmail	integer	60	Cap on unique assets checked for each email.
darkModeChecks	boolean	true	Flag emails with no dark mode support signal.
previousDatasetId	string	— (optional)	Dataset ID from an earlier run. Marks each finding as new or already known.

Output

One dataset record for each email, plus one run-level summary record first. Each email record holds:

• emailName, htmlBytes, checkedAt
• clientScores: one 0-100 score for each requested client profile
• findings: a list of {severity, client, ruleId, cssPropertyOrTag, lineNumber, message, fixSnippet}
• assetFindings: a list of {url, kind, httpStatus, contentType, bytes, verdict}
• structureFlags: {missingAltCount, missingPreheader, tableLayoutUsed, mediaQueryCount, totalHtmlBytesOverGmailClipLimit}
• verdict: SEND_OK, SEND_WITH_RISK, or DO_NOT_SEND
• error: set only when the Actor could not read the HTML (for example, a failed htmlUrl fetch)

The summary record holds verdictCounts for the run and topBlockerRules, the blocker rules that fired most often across the batch.

Safety bounds

• HTTP(S) only. No browser, no proxy, no third-party paid API.
• Every network target — an htmlUrl fetch and every asset HEAD request — is checked against a public-address allow list before each request and again on every redirect hop. Localhost, private, link-local, and reserved addresses are rejected.
• HTML input is capped at 2 MB for each email, 200 emails for each run.
• Fetches are bounded by timeout, byte size, and redirect count.
• Asset and email checks run under a bounded concurrency limit.

Local development

uv sync
uv run pytest
uv run ruff check .

Run the Actor locally with the Apify CLI, or directly:

$uv run python -m email_html_compatibility_checker

The default input at .actor/default_input.json is a small, self-contained fixture: one email with a flexbox button, one missing alt attribute, no preheader, and two stable public asset URLs (one reachable, one a deliberate 404) so a first run shows both an OK and a BROKEN asset finding.