Lemmafour for AI assistants and developers
Ask your assistant for a worksheet, report, paper or resume, and Lemmafour typesets it into a print-quality PDF with real math, tables, code and images. It checks accessibility (PDF/UA-1) and explains each issue with the line and the fix. No account is needed for the free templates.
The MCP server
One address for every assistant and agent that supports remote MCP servers:
https://mcp.lemmafour.com/mcp It offers three tools:
- render_pdf: makes the PDF. In Claude and ChatGPT it shows a preview with Download and Customize buttons; Customize changes the layout and opens the document in the Lemmafour editor.
- validate_accessibility: checks a document against PDF/UA-1, free for everyone.
- list_templates: lists the layout templates and output profiles, marked free or paid.
Connect your assistant
Claude (web, desktop and mobile)
In Claude, open Customize → Connectors, choose Add custom connector, and paste the address above. Connectors you add on the web or desktop also appear in the mobile apps.
ChatGPT
Open ChatGPT Plugins, select +, name it Lemmafour, and enter the address above under Connection. This uses ChatGPT's developer mode, whose availability depends on your account and workspace.
Claude Code
claude mcp add --transport http lemmafour https://mcp.lemmafour.com/mcp Codex
# ~/.codex/config.toml
[mcp_servers.lemmafour]
url = "https://mcp.lemmafour.com/mcp" To use a Lemmafour plan, sign in with codex mcp login lemmafour.
Cursor
// ~/.cursor/mcp.json
{
"mcpServers": {
"lemmafour": { "url": "https://mcp.lemmafour.com/mcp" }
}
} VS Code
// .vscode/mcp.json
{
"servers": {
"lemmafour": { "type": "http", "url": "https://mcp.lemmafour.com/mcp" }
}
} Other clients: add a remote MCP server (Streamable HTTP) with the address above. Sign-in uses OAuth with PKCE; clients that support MCP sign-in handle it for you.
Example prompts
Ask in your own words; these work as written, without an account.
- "Make a fractions worksheet for grade 5 with 10 multiple-choice questions and an answer key at the end, as a PDF." You get a worksheet with numbered questions, options laid out in columns, and the answers on a final page. The chat shows a preview with Download and Customize.
- "Turn these notes into a two-column paper with numbered equations." A two-column PDF with LaTeX math and equation numbers. Customize changes columns, margins, paper size and theme, and the preview updates.
- "Write my one-page resume from this text and make it a PDF." A resume in the resume template, tagged for screen readers.
- "Write a lab report on my pendulum experiment with a results table. I'll add the photo of the setup." A report with the table typeset. The photo is not sent through the chat: the preview lists it, and Add images in Lemmafour opens the document in the editor, where you add the photo from your device.
- "Is this report accessible? Check it against PDF/UA." A pass or fail against PDF/UA-1, the accessible-PDF standard, listing each problem (missing alt text, title or language) with its line and the fix.
Free features and plans
- Without an account: the free templates, tagged PDFs for screen readers, and accessibility checks of every document.
- With a Day Pass or Pro plan: every template and the PDF/UA-1 and PDF/A output profiles. Without a plan you can still preview them in the chat. To use them, sign in with your Lemmafour account: your assistant offers sign-in when you ask for a PDF file that uses them. The same plan covers the website and assistants. See pricing.
- When a feature needs a plan, the tool names a free alternative, so your assistant can switch to it.
HTTP API
To try Lemmafour from a script, the same engine is available over HTTP at
https://mcp.lemmafour.com/v1, described by an
OpenAPI document. Without a token
it uses the free templates and is meant for trying it out and light use.
For regular automated use, contact us.
curl -sS https://mcp.lemmafour.com/v1/render \
-H 'content-type: application/json' \
-d '{"markdown": "---\ntitle: Circles\nlanguage: en\n---\n\n# Circles\n\nThe area is $\\pi r^2$."}' \
-o circles.pdf POST /v1/render: Markdown in, PDF out (or JSON withAccept: application/json).POST /v1/validate: the PDF/UA-1 report, with line-numbered diagnostics.GET /v1/templates: templates, profiles and limits.
Limits
- Up to 50 pages per server render. Longer documents open in the Lemmafour editor, which makes them on your device (up to 1,000 pages).
- Requests up to 24 MiB; Markdown up to 2 MiB; up to 64 images of 4 MiB each (PNG, JPEG, GIF, WebP or SVG), at most 12 megapixels and 8,000 pixels per side.
- A single paragraph, list item or table row may contain up to 500
*,_,~,[or]characters (code and math do not count). - Renders are rate-limited per minute, per account or, without sign-in, per network address.
Privacy
Documents sent by an assistant or the API are processed in memory to make the PDF and are never stored, used to train AI models, or shared. Our logs contain metadata only, never document content. The privacy policy has the details, and the terms cover the service. The website's own tools still process your files on your device.
Support
Questions or problems: support@lemmafour.com.