A Model Context Protocol (MCP) server that gives your AI assistant the power to convert Markdown into 23+ professional document formats, analyze documents, repair broken Markdown, and convert across platforms β all without leaving your AI chat.
npm: https://www.npmjs.com/package/@xjtlumedia/markdown-mcp-server
You asked an AI to write 20 questions. It delivered. But now you're stuck:
| Pain Point | Manual Workflow | With AI Answer Copier |
|---|---|---|
| Extracting Q&A | 10β15 min copying each line | 2 seconds (auto-detect) |
| Formatting Math | 20 min fixing broken symbols | Instant (KaTeX support) |
| LMS Upload | 15 min manual CSV entry | 1-click export |
| Total Prep Time | ~45β60 minutes | < 1 minute |
Generating content takes seconds. Formatting it for the real world takes an hour. This MCP server eliminates that gap entirely.
Online web version (no install needed): ai-answer-copier.vercel.app
Test MCP tool in inspector:
Use actual MCP instance:
| # | Tool | Output | Description |
|---|---|---|---|
| 1 | harmonize_markdown |
.md |
Standardize and normalize Markdown syntax (ATX headers, - list markers, fenced code blocks) without changing meaning |
| 2 | convert_to_txt |
.txt |
Strip all formatting to produce plain text |
| 3 | convert_to_html |
.html |
Full styled HTML document with GFM tables, KaTeX math, and embedded stylesheet |
| 4 | generate_html |
HTML string | Generate self-contained HTML with all styles inlined (read-only, no file I/O) |
| 5 | convert_to_pdf |
.pdf |
Print-ready PDF via headless Chromium with full KaTeX math and syntax highlighting |
| 6 | convert_to_docx |
.docx |
Microsoft Word document with styled headings, lists, and code formatting |
| 7 | convert_to_latex |
.tex |
LaTeX source code with \section, list environments, and native math pass-through |
| 8 | convert_to_rtf |
.rtf |
Rich Text Format for legacy word processors and email clients |
| 9 | convert_to_csv |
.csv |
Extract GFM pipe-tables to comma-separated values |
| 10 | convert_to_json |
.json |
Structured JSON representation with sections, headings, lists, code blocks, and tables |
| 11 | convert_to_xml |
.xml |
Well-formed XML with <?xml?> declaration and structured elements |
| 12 | convert_to_xlsx |
.xlsx |
Excel spreadsheet β each Markdown table becomes a sheet |
| 13 | convert_to_image |
.png |
Full-page PNG screenshot via headless Chromium |
| 14 | convert_to_md |
.md |
Export Markdown to file, optionally harmonizing formatting first (harmonize=true) |
| 15 | convert_to_email_html |
.html |
Email-optimized HTML with all styles inlined, compatible with Outlook, Gmail, Apple Mail |
| # | Tool | Target Platform | Description |
|---|---|---|---|
| 16 | convert_to_slack |
Slack | mrkdwn format β bold as *, links as <url|text>, headers as bold text |
| 17 | convert_to_discord |
Discord | Styled bold/underline headers, preserved code blocks |
| 18 | convert_to_jira |
JIRA | Wiki markup β h1./h2., {code} blocks, [text|url] links |
| 19 | convert_to_confluence |
Confluence | Wiki markup with {info}, {note} panels and {code:language=x} syntax |
| 20 | convert_to_asciidoc |
AsciiDoc | = headers, ---- code blocks, url[text] links, image:: directives |
| 21 | convert_to_rst |
reStructuredText | Underlined headers, .. code-block:: directives, RST reference syntax |
| 22 | convert_to_mediawiki |
MediaWiki | == headers, '''bold''', <syntaxhighlight> tags, {\ wikitable |
| 23 | convert_to_bbcode |
BBCode | [b], [i], [code], [url], [img] tags for phpBB/vBulletin forums |
| 24 | convert_to_textile |
Textile | Markup for Redmine, Basecamp, and CMS platforms |
| 25 | convert_to_orgmode |
Emacs Org Mode | * headers, #+BEGIN_SRC/#+END_SRC code, [[url][text]] links |
| # | Tool | Description |
|---|---|---|
| 26 | html_to_markdown |
Convert HTML content (full document or fragment) back to Markdown |
| # | Tool | Description |
|---|---|---|
| 27 | repair_markdown |
Fix broken Markdown from LLM output or copy-paste β unclosed code fences, broken tables, stray emphasis markers, missing heading spaces, broken links |
| 28 | lint_markdown |
Lint and report issues as JSON array with line number, severity, rule name, message, and fixable flag |
| # | Tool | Description |
|---|---|---|
| 29 | extract_code_blocks |
Extract all code blocks with language, content, and start/end line numbers |
| 30 | extract_links |
Extract all links and images with text, URL, line number, and type (inline, reference, image, autolink) |
| 31 | generate_toc |
Generate Markdown Table of Contents from headings with configurable max_depth (1β6) |
| 32 | analyze_document |
Comprehensive statistics β line/word/character/paragraph/sentence counts, heading/code/table/link/image counts, estimated reading time |
| 33 | extract_structure |
Bird's-eye document architecture β statistics, heading outline, code block summary, link summary |
Pre-configured prompt templates that orchestrate multiple tools:
| Prompt | Description |
|---|---|
convert-document |
Convert Markdown to any supported format (PDF, DOCX, HTML, LaTeX, CSV, JSON, XML, XLSX, RTF, PNG, TXT, MD) |
extract-tables |
Extract all tables from Markdown and export as CSV or XLSX |
format-for-sharing |
Harmonize formatting, then convert to both PDF and HTML for distribution |
analyze-and-repair |
Lint β Repair β Re-lint to confirm all issues resolved |
convert-for-platform |
Convert Markdown to any platform format (Slack, Discord, JIRA, Confluence, AsciiDoc, RST, MediaWiki, BBCode, Textile, Org Mode) |
document-overview |
Full overview β statistics, TOC, code blocks, and links in one report |
| Resource URI | Description |
|---|---|
markdown-formatter://supported-formats |
Complete JSON list of all 23+ output formats with tool names and types |
markdown-formatter://conversion-guide |
Guide for choosing the right output format based on your use case |
No installation needed β just configure your MCP client:
{
"mcpServers": {
"ai-answer-copier": {
"command": "npx",
"args": ["-y", "@xjtlumedia/markdown-mcp-server"]
}
}
}npm install -g @xjtlumedia/markdown-mcp-serverThen configure:
{
"mcpServers": {
"ai-answer-copier": {
"command": "markdown-mcp-server"
}
}
}
Edit %APPDATA%\Claude\claude_desktop_config.json (Windows) or ~/Library/Application Support/Claude/claude_desktop_config.json (macOS):
{
"mcpServers": {
"ai-answer-copier": {
"command": "npx",
"args": ["-y", "@xjtlumedia/markdown-mcp-server"]
}
}
}Restart Claude Desktop. You'll see a π icon β all 33 tools are now available.
Add to .vscode/mcp.json or VS Code settings:
{
"servers": {
"ai-answer-copier": {
"command": "npx",
"args": ["-y", "@xjtlumedia/markdown-mcp-server"]
}
}
}Use the same npx command pattern above in your client's MCP configuration.
{
"mcpServers": {
"markdown-formatter": {
"command": "node",
"args": [
"YOUR_PATH_TO_PROJECT/mcp-server/dist/mcp-server/src/index.js"
]
}
}
}Replace YOUR_PATH_TO_PROJECT with the absolute path to this folder.
A hosted HTTP endpoint is available for browser-based and remote integrations:
https://ai-answer-copier.vercel.app/api/mcp
Test with MCP Inspector:
npx @modelcontextprotocol/inspector https://ai-answer-copier.vercel.app/api/mcpOnce connected, just talk to your AI naturally:
"Generate 10 physics questions about Newton's Laws and export them as a Kahoot CSV." β AI uses
convert_to_csv
"Convert this markdown into a Word document and save it to my Desktop." β AI calls
convert_to_docxwithoutput_path
"Take these lecture notes and produce a print-ready PDF." β AI calls
convert_to_pdfwith Puppeteer rendering, KaTeX math, and syntax highlighting
"Turn this into a Moodle-compatible XML quiz bank." β AI calls
convert_to_xml
"Format this for Slack and send it to the team channel." β AI calls
convert_to_slack
"Check this Markdown for issues and fix them." β AI calls
lint_markdownthenrepair_markdown
"Give me an overview of this document β word count, structure, all links." β AI chains
analyze_document+generate_toc+extract_links
Export to PDF, DOCX, HTML, LaTeX, CSV, JSON, XML, XLSX, RTF, PNG, TXT, Markdown β plus Slack, Discord, JIRA, Confluence, AsciiDoc, RST, MediaWiki, BBCode, Textile, Org Mode, and Email HTML.
Full support for KaTeX math expressions and syntax-highlighted code blocks. Your $\sqrt{x^2 + y^2}$ and Python snippets survive every conversion perfectly.
Binary formats (PDF, DOCX, XLSX, PNG) don't dump raw base64 into chat. The server returns actionable guidance so the AI knows to save to a file path.
Lint Markdown, auto-repair broken formatting, extract code blocks and links, generate Table of Contents, and get comprehensive document statistics β all without leaving the chat.
Tables, task lists, strikethrough, autolinks β all parsed correctly via remark-gfm.
PDF and image exports use headless Chromium for pixel-perfect rendering with full CSS styling, KaTeX math, syntax highlighting, and print-optimized layouts.
Zero runtime dependencies on external APIs. Everything runs locally. Your data never leaves your computer.
βββββββββββββββββββββββββββ
β AI Client β
β (Claude, Copilot, β
β Cursor, Windsurf) β
ββββββββββββ¬βββββββββββββββ
β MCP Protocol (stdio / HTTP)
βΌ
βββββββββββββββββββββββββββ
β AI Answer Copier β
β MCP Server (33 tools) β
β β
β βββ Conversion ββββββββββ
β β remark/rehype, docx, ββ
β β puppeteer, xlsx, ββ
β β CSV/JSON/XML/RTF/ ββ
β β LaTeX/TXT parsers ββ
β βββββββββββββββββββββββββ
β βββ Platform ββββββββββββ
β β Slack, Discord, JIRA,ββ
β β Confluence, AsciiDoc,ββ
β β RST, MediaWiki, ββ
β β BBCode, Textile, Org ββ
β βββββββββββββββββββββββββ
β βββ Analysis ββββββββββββ
β β Code blocks, links, ββ
β β TOC, stats, structureββ
β βββββββββββββββββββββββββ
β βββ Repair ββββββββββββββ
β β Lint, auto-repair ββ
β βββββββββββββββββββββββββ
β βββ Import ββββββββββββββ
β β HTML β Markdown ββ
β βββββββββββββββββββββββββ
βββββββββββββββββββββββββββ
This release hardens both the local (npx) and remote Vercel MCP transports with 8 logic fixes:
| # | Fix | Area |
|---|---|---|
| 1 | Browser not closed on error β try/finally around browser.close() so Chromium never leaks on PDF/PNG failures |
Both transports |
| 2 | generate_html missing KaTeX CSS β restored viewport meta, KaTeX stylesheet (SRI hash), and full inline styles so math renders correctly |
Vercel /api/mcp |
| 3 | convert_to_html incomplete styles β added SRI integrity hash, table borders, and blockquote styling to match local output |
Vercel /api/mcp |
| 4 | Session map memory leak β instances Map now evicts stale sessions after 30 minutes (SESSION_TTL_MS); McpInstance tracks lastUsed |
Vercel /api/mcp |
| 5 | keepAlive interval not cleared on normal SSE stream end β added inner try/finally that always calls clearInterval(keepAlive) |
Vercel /api/mcp |
| 6 | _initialized SDK-internal access β replaced bare @ts-ignore with a typeof guard so the server fails gracefully after SDK upgrades |
Vercel /api/mcp |
| 7 | No input size validation β added 1 GB hard limit (MAX_INPUT_BYTES) to all tool calls to prevent runaway CPU/memory on oversized payloads |
Both transports |
| 8 | DELETE method creates new session β handler now returns 200 immediately on DELETE, preventing phantom session creation |
Vercel /api/mcp |
- PDF / PNG require local Chromium β
convert_to_pdfandconvert_to_imagelaunch a headless browser. This works fine withnpxbut is unavailable on Vercel's free tier (no bundled Chromium). Set thePUPPETEER_EXECUTABLE_PATHenvironment variable to point to your local Chrome if auto-detection fails. - Chromium version on Vercel β the remote endpoint uses
@sparticuz/chromium-minwith a pinned tar at v131.0.1. If Vercel bumps the Lambda runtime you may need to update theCHROMIUM_TAR_URLenv var. - 1 MB input limit β all tool calls reject inputs larger than 1 MB. Split large documents into sections before converting.
- Session TTL (remote only) β sessions on the Vercel endpoint expire after 30 minutes of inactivity. Long-running stateful workflows should issue periodic keep-alive calls.
# Clone the repository
git clone https://github.com/XJTLUmedia/AI_answer_copier.git
# Web app
npm install
npm run dev
# MCP server
cd mcp-server
npm install
npm run dev # hot reload
npm run build # production build
npm run inspector # test with MCP Inspector| Component | Technology |
|---|---|
| MCP SDK | @modelcontextprotocol/sdk |
| Markdown Parser | unified + remark + rehype |
| Math Rendering | remark-math + rehype-katex |
| Word Export | docx |
| Excel Export | xlsx |
| PDF/PNG Export | puppeteer |
| Schema Validation | zod |
| Runtime | Node.js (ESM) |
We welcome contributions! Whether it's a new export format, a bug fix, or documentation improvement β please consider making a pull request.
- Fork the repository
- Create a feature branch
- Submit a pull request
MIT Β© XJTLUmedia
Built by educators, for educators.
Reclaim 5 hours of your week. Let the AI handle the formatting.