Unstructured

👀 Have you tried Unstructured Transform MCP yet? It's an MCP that drops into the agent you already work in, Claude Code, Cursor, Codex, in about three minutes. After that, parsing stops being something you think about. You point your agent at a file and say what you want in plain English. "Parse this and give me the tables." Transform reads the file, picks the right strategy, handles the OCR and chunking, and hands structured data back in the same session. You don't install a Python package, tune API parameters, or benchmark a parsing strategy. You describe the outcome and it works out the how. It holds up on the files that usually break things too: scanned PDFs, handwriting, dense tables, 65+ formats. Point it at your ugliest file and see what comes back: https://lnkd.in/eHes2eb9

We’re headed to Austin for Fed Supernova 🤠 Come find Unstructured at the Dual-Use Startup Crawl next Tuesday! We’ll have fun swag and great conversations about getting complex, unstructured data ready for production AI systems. 📍 Austin, TX 📅 Tuesday, August 18 ⏰ 5 - 9 PM See you there!

If you've ever tried to turn a messy PDF, scanned contract, or complex spreadsheet into something an LLM can actually use, you know the drill: pick a partitioning strategy, guess at chunking params, write glue code, run it, get mediocre results, repeat. Half a day gone, and all you have is a script held together with hope. We built Unstructured Transform MCP to fix that. It's an MCP server that lives inside Claude, Codex, Cursor, VS Code, or wherever your agents already work. Describe what you want in plain English → get structured output back → get the actual working code (Python or curl) to productionize it. Not templates. Not pseudocode. Under the hood: - Ingests 60+ formats (PDF, DOCX, PPTX, XLSX, HTML, scanned images) - One pipeline for partitioning, enriching, chunking, and embedding — not four tools to babysit - Outputs JSON, Markdown, HTML, or plain text — ready for your vector store or agent What it unlocks: - Raw files → AI-ready data, no ETL expertise required - Learn partitioners, chunkers & embedders by using your own data - Walk away with sample code you can drop into production The old way to "try" a document AI product means reading docs and guessing at configs. We wanted trying it to be the real work — point to your files, describe what you need, see it run, grab the code, done. Minutes, not half a day. Try it: https://lnkd.in/eHes2eb9 Docs: https://lnkd.in/ekQ_Tvy9

Unstructured's Email Notification Channels alert your team the moment a workflow job completes, fails, or hits a snag. No babysitting required. 😎 Configure alerts at the workspace level for broad visibility, or drill down to individual workflows for surgical precision. Setup takes just two API calls: one to create the channel, one to verify it. Let your inbox do the watching so you can focus on what actually matters. ✨ Learn more: https://lnkd.in/eXCdaXDu

📚 You have a stack of customer order forms, real estate flyers, or care guides. Same layout, different content every time. Normally, extracting the fields you need means writing custom parsing logic for each format. Unstructured Transform MCP skips that 💅 Describe what you want in plain language, and the agent can suggest a schema from the file, draft one from your instructions, or extract against a schema you provide. Working with a batch of similar documents? Ask for one consistent structure, and it keeps that schema steady across every file. Learn more: https://lnkd.in/eNez7JqR

Go get some swag!!!

You asked, we listened 👀🛍️ We’re so excited to finally launch the Unstructured Shop!! 🥳 After hearing how much you all love our swag, we figured it was time to make it a little easier to get your hands on it 😉 This has been a long time coming, and we’re just getting started. We’ll be adding even more items over the next few weeks, so keep an eye out for new drops ✨ Shop now 👉 https://lnkd.in/efjaVGdy

Back to Tampa we go! 🌴 We’re excited to join Carahsoft at #DoDIIS26 on Monday. Be sure to stop by Kiosk 2 in the Carahsoft Pavilion to learn how Unstructured helps organizations turn complex documents into clean, AI-ready data. 📍 Kiosk 2 at the Carahsoft Pavilion 📆 August 10 ⏰ 8:00 AM–5:00 PM See you there!

Always love bringing the Unstructured + IBM community together 🐝 🫶 Huge thanks to everyone who joined, asked thoughtful questions, and shared how they’re thinking about agents, MCP, and production AI systems.

A developer reads your docs once and moves on, but a model reads your tool description every time it decides whether to call you, with a finite context window and no patience for ambiguity. Building Unstructured Transform MCP made us rethink tool descriptions, error messages, response payloads, and even how many tools an agent should have access to. Check out what it takes to build software for the agentic era 👇