What Is PDF123: A Free PDF Toolkit You Can Also Self-Host
PDF123 is a free browser-based PDF toolkit: all 56 tools, from merge and compress to OCR and encryption, run without an account. The same operations are exposed over REST, MCP, and the pdfx CLI, or you can self-host the stack and keep files on your own network.
Most PDF work is small: merge a few files into one, compress a scan until it fits in an email, pull the text out of a scan, put a password on a contract. Small, but it comes back many times a year, and each time you pick a tool again. The sites you find usually fail in one of two ways: they cap the number of jobs, they want an account, or they take your file somewhere you cannot account for.
PDF123 is built for that recurring small job: a free site, no account, download when it finishes. The same toolkit has two other entry points, for people who need to process in bulk or who cannot send a file off their network.
Five groups, 56 actions
The 56 tools on the site fall into five groups, named the same way as the home page navigation: Page organizing, Format conversion, Content editing, Security and encryption, and Advanced tools. Every group is concrete actions, so nothing about the internal structure of a PDF is a prerequisite.
| Group | What is in it | Count |
|---|---|---|
| Page organizing | Merge, split, rotate, reorder, crop, drop blank pages, add page numbers | 15 |
| Format conversion | PDF to and from Word, Excel, Markdown, images, and HTML, plus ebooks and Office documents | 16 |
| Content editing | Compress, repair, flatten forms, edit metadata, extract images, decompress streams | 14 |
| Security and encryption | Add password, remove password, watermark and stamp, redact text, sanitize active content, validate signatures | 8 |
| Advanced tools | OCR to Markdown, read document info, list embedded JavaScript | 3 |
Above: the five tool groups by purpose, 56 in total. The grouping and the counts come from the site's own tool catalog and match the home page.
The frequently used ones have direct entries on the home page. In organizing, the most-clicked are Merge PDF and Split PDF; in content editing, Compress PDF and Repair PDF; to encrypt, Add Password; for scans, OCR to Markdown.
Three ways to call it, one processing path
The browser is the shortest path: upload, set parameters, process, download, no account needed. For one or two files, every extra step is pure overhead.
Bulk work goes through the API or a CLI. The same operations are served as REST endpoints, an MCP endpoint, and a pdfx command line, so parameters you tuned in the browser move into code as they are, and hundreds of files run in one pass without re-reading the docs per operation. Anonymous calls need no account and no API key. Explanations and examples are on the Developers page; the four client types are compared in Same operation, four clients.
Self-hosting deploys the backend and the portal onto your own servers, so files never leave your network. The hosted site and your instance run the same implementation; what changes is who operates it: upgrades, backups, and capacity are yours to manage, and the cost is spelled out in What self-hosting actually buys you, and what it costs. Deployment is covered under Self-host.
The decision compresses to one line: one file goes through the browser; the same class of file every week moves to a script; a file that may not leave the internal network gets a self-hosted instance.
What keeps it free
Every tool works without signing in, no operation is marked paid, and there is no pricing page. The real limit is anonymous rate limiting rather than a quota wall: requests that come too fast are held back for a while, and work again after the cooldown. There is no “you have used up today's jobs” that locks the door.
That limit applies to request rate, not to file count or page count, and it does not require signing up first to lift. The trade-off is that at peak times you may need to wait a moment and retry.
What it deliberately does not do
OCR to Markdown returns Markdown text, not a searchable PDF with a hidden text layer, and it does not reconstruct layout. The tool name says so, and expecting a full-text-searchable PDF back from it will disappoint.
Watermarks and stamps support only Helvetica with WinAnsi encoding, so non-Latin text is not rendered correctly. The limits that get misread most often are collected below; the tool pages and the API docs use the same wording.
| Common misreading | What actually happens |
|---|---|
| OCR returns a searchable PDF | It returns Markdown text, does not reconstruct layout, and adds no hidden text layer |
| Watermarks and stamps take any script | Helvetica with WinAnsi encoding only; non-Latin text is rendered wrong |
| Remove certificate signature strips the signature | The endpoint currently returns the uploaded file unchanged, signature included |
| There is a free tier with paid plans | There is no pricing page; limits come from anonymous rate limiting, not per-job billing |
Above: four frequent misreadings against what actually happens. The first two carry the same note on their tool pages.
It is also not a substitute for a forensic suite, a print RIP, or a certified archival workflow. Output is best-effort conversion and editing, so keep your original file. Encryption, sanitizing, and signature removal are security-facing tools for documents you already have the right to modify; they are not a way around someone else's access control.
Where to start
Open pdf123.xyz and pick a tool by what you need to do. Who maintains it, how uploads are handled, and how to make contact are all on the About page.
To wire it into your own pipeline, three posts are the natural next reads: Same operation, four clients puts the browser form, curl, MCP, and pdfx side by side; pdfx CLI: one catalog, called from your terminal covers the command line; What self-hosting actually buys you, and what it costs covers the cost of moving the whole stack onto your own network.