# LaunchGrade — full context for AI agents

> LaunchGrade is a free, open-source, zero-dependency launch audit from FastPress Labs. It grades a public page across 54 checks in 7 weighted categories and returns prioritized fixes. No account, no telemetry, no paid API. Node.js 18.17+.

Canonical URL: https://fastpresspages.github.io/launchgrade/
Source: https://github.com/fastpresspages/launchgrade
License: Apache-2.0
Contact: meetcode99@gmail.com
Last updated: 2026-10-23

## How to run it

```bash
npx -p github:fastpresspages/launchgrade launchgrade https://example.com
```

Write all report artifacts:

```bash
launchgrade example.com --out reports
# report.html, report.json, report.md, badge.svg, fix-plan.md
```

Compare a new scan against a saved baseline (one artifact with the full before/after delta):

```bash
launchgrade example.com --compare reports/report.json --out reports/after
```

`report.json` is documented by a JSON Schema (`schema/report.schema.json`, draft 2020-12; each file carries a `$schema` URL). Additive fields may appear at any release; breaking changes require a major bump.

Use in CI:

```yaml
- uses: fastpresspages/launchgrade@main
  with:
    url: https://example.com
    fail-under: 80
```

Use as an MCP server:

```json
{
  "mcpServers": {
    "launchgrade": {
      "command": "npx",
      "args": ["-y", "-p", "github:fastpresspages/launchgrade", "launchgrade-mcp"]
    }
  }
}
```

MCP tools: `grade_launch` (url, format markdown/json, max_fixes), `get_fix_plan` (url), `list_checks` (no arguments).

## Scoring

Each check is pass (1), warn (0.5), fail/error (0), or info (excluded). Categories are normalized to 100 and weighted:

| Category | Weight |
| --- | ---: |
| Conversion | 25% |
| Trust | 15% |
| Technical | 20% |
| SEO | 15% |
| Social sharing | 10% |
| Accessibility | 5% |
| AI-agent readiness | 10% |

Grades: A+ >=95, A >=90, A- >=85, B+ >=80, B >=75, B- >=70, C+ >=65, C >=60, C- >=55, D >=40, F <40.
Labels: Launch-ready >=90, Almost there >=80, Needs polish >=70, Risky launch >=60, Not ready <60.

## Full checklist

### Conversion (25%)
- conv.email-capture — Has an email capture
- conv.cta — Has a specific call to action
- conv.social-proof — Shows social proof
- conv.friction — Keeps signup friction low
- conv.contact — Offers a human contact path
- conv.value-prop — States a clear value proposition
- conv.risk-reversal — Reverses the risk of clicking

### Trust (15%)
- trust.privacy — Links a privacy policy
- trust.terms — Links terms of service
- trust.about — Explains who is behind the page
- trust.social-presence — Links a real social profile
- trust.security-txt — Publishes a security contact
- trust.copyright — Shows a current footer

### Technical (20%)
- tech.https — Serves over HTTPS
- tech.status — Returns a healthy status
- tech.speed — Loads the HTML quickly
- tech.compression — Compresses the response
- tech.cache — Sets cache headers
- tech.viewport — Is mobile-ready
- tech.indexable — Allows search indexing
- tech.canonical — Declares a canonical URL
- tech.lang — Declares the page language
- tech.favicon — Has a favicon
- tech.mixed-content — Avoids mixed content
- tech.analytics — Measures visits

### SEO (15%)
- seo.title — Has a well-sized title
- seo.description — Has a meta description
- seo.h1 — Has exactly one H1
- seo.heading-order — Uses a sane heading outline
- seo.alt — Describes its images
- seo.sitemap — Publishes a sitemap
- seo.robots — Publishes robots.txt
- seo.schema — Adds structured data
- seo.internal-links — Links its own pages
- seo.text-ratio — Has readable copy in the HTML

### Social sharing (10%)
- social.open-graph — Has a complete Open Graph card
- social.twitter — Has a Twitter/X card
- social.og-image-alt — Describes the share image

### Accessibility (5%)
- a11y.form-labels — Labels its form fields
- a11y.button-names — Gives buttons accessible names
- a11y.link-text — Writes descriptive link text
- a11y.landmarks — Uses landmark elements
- a11y.zoom — Allows pinch-zoom
- a11y.skip-link — Offers a skip link

### AI-agent readiness (10%)
- agent.llms-txt — Publishes llms.txt
- agent.llms-full — Offers a full-context file
- agent.robots-ai — Lets AI crawlers in
- agent.agents-md — Speaks to coding agents
- agent.markdown — Serves markdown to agents
- agent.mcp — Exposes a machine-readable API
- agent.faq — Answers buying questions
- agent.semantic-html — Uses semantic HTML
- agent.no-js-copy — Works without JavaScript

## What it does not do

- It does not execute JavaScript or render the page in a browser.
- It does not log in, bypass paywalls or scan private pages.
- It does not measure Core Web Vitals; use Lighthouse or CrUX for that.
- It does not guarantee rankings, conversions or revenue.
- It does not use AI to invent findings. Every check is deterministic code.

## Real sample

Two dated scans of https://fastpresspages.github.io/YouLauncher/ on 23 October 2026: the original page returned B- 71/100 (conversion 93, trust 38, technical 91, SEO 79, social 20, accessibility 100, AI-agent readiness 50), and the rebuilt page returned A+ 100/100 with one warning (48 scored: 47 pass, 1 warn, 0 fail; technical 98; the warning is the deliberate absence of analytics). Reports: https://fastpresspages.github.io/launchgrade/samples/youlauncher-before/report.html and https://fastpresspages.github.io/launchgrade/samples/youlauncher/report.html. These are dated snapshots, not claims about future performance.

## Free human audit

https://fastpresspages.github.io/launchgrade/#audit — send the URL, what you are launching and what is stuck. A human reads it and replies. Also open to honest collaborations, sponsorships and deals: meetcode99@gmail.com.