Toolora MCP Server

Extract all plain text from a PDF file. Accepts the file as a base64-encoded string. Returns the text content, page count, word count, and character count. Useful for reading documents, research papers, invoices, or any PDF without needing to open it.

Merge 2 to 20 PDF files into a single PDF document. Each file is provided as a base64-encoded string. Returns the merged PDF as a base64 PNG. Ideal for combining reports, invoices, chapters, or any multi-file PDF workflow.

Generate a QR code image from any text or URL. Returns a base64-encoded PNG. Supports custom size and error correction level. Useful for creating shareable links, contact cards, payment addresses, or any scannable code.

Convert a CSV string into a JSON array of objects. The first row is treated as column headers by default. Handles quoted fields, embedded commas, and CRLF line endings. Useful for processing spreadsheet exports, data files, and tabular data.

Convert a JSON array of objects into a CSV string. Column names are inferred from all object keys. Handles special characters, embedded commas, and quotes with proper escaping. Returns a CSV with CRLF line endings.

Extract text from an image using OCR (Optical Character Recognition). Accepts the image as a base64-encoded string. Returns extracted text, confidence score, and word count. Useful for reading scanned documents, screenshots, handwritten notes, or any image where text needs to be extracted without a vision model.

Extract all plain text from a Microsoft Word (.docx) file. Accepts the file as a base64-encoded string. Returns the full text, paragraph count, word count, and character count. Works with files from Microsoft Word, Google Docs, and LibreOffice.

Parse an Excel file (.xlsx, .xls, .csv, .ods) and return all sheets as JSON arrays of objects. Column headers from the first row become object keys. Returns a map of sheet name to row array, plus total row count.

Transcribe speech from an audio file to text using OpenAI Whisper (gpt-4o-mini-transcribe). Accepts the audio as a base64-encoded string. Returns the full transcript, word count, and character count. Supported formats: flac, m4a, mp3, mp4, mpeg, mpga, oga, ogg, wav, webm. Maximum file size: 25MB. Great for transcribing meeting recordings, podcasts, voice notes, interviews, or any spoken audio.

Fetch and extract content from a web page that requires JavaScript rendering (SPAs, dynamically loaded content). Unlike read_url which uses simple HTTP fetch, this tool uses a headless browser to render the page first. Returns clean text or markdown. Use for React/Next.js/Angular/Vue apps, product pages, news sites behind JS paywalls, or any page where read_url returns empty content.

Count words, characters, sentences, paragraphs, and unique words in a block of text. Returns reading time (200 wpm), speaking time (130 wpm), Flesch readability score with label (Very Easy → Very Difficult), and top keyword frequency excluding common stop words. Use when you need document stats, content length validation, readability checks, or keyword density analysis.

Fetch a public web page and return its main readable content as plain text, along with the title, author, and excerpt. Uses Mozilla Readability to strip navigation, ads, and boilerplate. Great for reading articles, documentation, or any page you want to process without a browser.

Generate a PDF document from plain text or markdown-style content. Supports headings (lines starting with #), bullet lists, and paragraphs. Returns the PDF as a base64-encoded string. Ideal for packaging Claude's output — reports, summaries, plans — into a shareable PDF file.

Convert markdown text into a clean, complete HTML document. Supports headings, bold, italic, code blocks, inline code, bullet lists, numbered lists, blockquotes, horizontal rules, and links. Returns a full HTML page ready to save, share, or embed. Great for turning Claude's markdown output into publishable web content.

Store any text content and get back a public URL at toolora.dev/p/[id] that anyone can visit to read it. Perfect for sharing Claude's output — reports, code, analysis — as a clean link without needing a file or account. Content expires after 7 days by default.

Save any text under a named key so you can recall it later — across sessions and conversations. Works like a persistent notepad: give it a key (e.g. 'project-spec', 'user-prefs', 'meeting-notes') and it's retrievable any time with recall_memory. Keys are global and upserted — saving to the same key overwrites previous content. Expires in 30 days by default.

Retrieve text you previously saved with save_memory. Pass the same key you used when saving. Returns the stored content plus when it was saved and when it expires. Returns an error if the key doesn't exist or has expired.