Skip to main content
Glama

Preflight MCP

Email campaign QA for AI agents. Preflight is a Model Context Protocol (MCP) server that gives AI assistants such as Claude a set of precise pre-send checks for marketing emails. Instead of working through a manual QA checklist, you ask:

"Preflight this email before we send it."

The agent decides which checks to run, calls Preflight's tools, and explains what is wrong and how to fix it.

Try it in your browser: the same checks as a single web page. Nothing is installed and nothing you paste is uploaded.

Why this exists

Pre-send QA on a busy email programme is slow and repetitive: broken links, missing UTM parameters, unrendered personalisation tags, missing alt text, a footer without an unsubscribe link. Mistakes cost revenue, data quality and compliance.

Major marketing platforms now ship MCP servers so AI agents can work inside them. Preflight adds the missing QA layer: deterministic, testable checks that any MCP-compatible agent can call, with an optional connector to Salesforce Marketing Cloud.

Related MCP server: Email Verifier MCP Server

What it checks

Tool

What it does

preflight_email

Full pre-send review (content, UTMs and compliance) with a verdict and score

check_email_content

Empty, javascript: or staging links; http links and images; missing alt text; unrendered merge tags (SFMC, Mailchimp, Handlebars/Liquid); Gmail clipping size; spam-trigger phrases; lang, <title> and preheader

validate_utm_links

Required UTM parameters on every tracked link, enforced values (e.g. utm_medium=email), lowercase naming, consistent utm_campaign. Unsubscribe, legal and mailto: links are skipped

check_email_compliance

Unsubscribe or preference link, postal address, privacy link and consent wording, by region (EU/GDPR, US/CAN-SPAM or global)

check_subject_line

Subject and preheader length, capitals, exclamation marks, spam phrases, merge tags

summarise_campaign_results

Open rate, CTR, click-to-open, bounce and unsubscribe rates from a CSV export, with flags for weak campaigns

preflight_sfmc_email

Fetches an email from Salesforce Marketing Cloud Content Builder by asset ID and runs the full review

It also includes a preflight_review prompt that sets the agent up as a QA lead.

Example

Running preflight_email on examples/broken_email.html returns FAIL with 5 blocking issues and 9 warnings, including:

  • Link "blog" points to a staging address and has no UTM parameters

  • Link "download the slides" has no destination (href="#")

  • Link "Click here" uses utm_medium=Email instead of email, and a campaign name with a space

  • Possible unrendered SFMC merge tag %%FirstName%%

  • Logo image has no alt text and loads over http

  • No unsubscribe or preference-centre link

The corrected version, examples/clean_email.html, returns PASS with a score of 100.

Quick start

Requires Python 3.10 or later.

git clone https://github.com/vibhu-sharma-09/preflight-mcp.git
cd preflight-mcp
python -m venv .venv
source .venv/bin/activate          # Windows: .venv\Scripts\activate
pip install -e ".[dev]"
pytest                              # 15 tests

Connect it to Claude Desktop

Open Claude Desktop, go to Settings > Developer > Edit Config, and add Preflight to claude_desktop_config.json:

{
  "mcpServers": {
    "preflight": {
      "command": "/full/path/to/preflight-mcp/.venv/bin/preflight-mcp"
    }
  }
}

On Windows the command is C:\\full\\path\\to\\preflight-mcp\\.venv\\Scripts\\preflight-mcp.exe. Restart Claude Desktop; the Preflight tools appear in the tools menu. Paste an email's HTML and ask Claude to preflight it.

Optional: connect to Salesforce Marketing Cloud

preflight_sfmc_email reads emails straight from Content Builder. An SFMC admin creates an Installed Package with an API Integration (server-to-server) component and read access to Content Builder, then you add the credentials to the server's environment:

"preflight": {
  "command": "/full/path/to/preflight-mcp/.venv/bin/preflight-mcp",
  "env": {
    "SFMC_SUBDOMAIN": "your-tenant-subdomain",
    "SFMC_CLIENT_ID": "...",
    "SFMC_CLIENT_SECRET": "...",
    "SFMC_ACCOUNT_ID": "optional business unit MID"
  }
}

The connector follows Salesforce's documented REST endpoints and is covered by mocked tests; try it in a sandbox business unit first. Use it only with your organisation's approval.

Browser version

docs/index.html runs the same rules in JavaScript, entirely in the browser, for teams that can't install software or connect an AI agent. It is published with GitHub Pages and also works offline: download the file and open it. To rebuild it after changing the samples, run python web/build.py.

How it is built

  • src/preflight/checks.py: the checks as plain Python functions, independent of MCP and easy to test

  • src/preflight/sfmc.py: Salesforce Marketing Cloud connector (OAuth client credentials, Content Builder asset API)

  • src/preflight/server.py: exposes the checks as MCP tools using the official MCP Python SDK

  • tests/: unit tests, including a mocked SFMC API

  • web/template.html and web/build.py: source for the browser version in docs/

Compliance checks are heuristics that flag likely gaps; final sign-off stays with a person.

Roadmap

  • Rendering checks across email clients

  • Link status checks (detect 404s)

  • Brand rules from a config file (approved domains, tone, banned words)

  • A multi-agent workflow: a planning agent builds a campaign through a platform's MCP server, and Preflight approves it before send

Licence

MIT

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      A
      quality
      C
      maintenance
      ▎ Check an email before an AI agent sends it: send / ask a human / block — with machine reason codes and prompt-injection detection.
      2
      2
      MIT
    • F
      license
      Not graded
      quality
      C
      maintenance
      Enables AI agents to verify email addresses in real-time, checking syntax, DNS, MX records, and SMTP handshake, returning verdicts and deliverability scores without sending actual emails.
      -
    • A
      license
      Not graded
      quality
      B
      maintenance
      Enables AI agents to check email HTML for delivery-breaking problems before sending, returning structured diagnostics with rule id, severity, line and column, along with a tool that lists every available rule. It runs with no filesystem, network or environment access, so agents can validate markup without gaining new permissions.
      15 npm
      5
      MIT
    • A
      license
      Not graded
      quality
      B
      maintenance
      Lets AI agents connect a user's database, draft marketing campaigns in their brand, preview emails, and send tests while the user retains final send control.
      49 npm
      MIT