Free browser tools

Matrix free tools handle small jobs on a phone or computer without an account, card, or Matrix computer. Open a tool, choose a local file or enter text, run it, then copy or download the result. The directory lists only tools that have a working browser action.

Privacy and processing

Text and files are processed in your browser. Your input is not uploaded to Matrix: the site does not save the source file or result to a Matrix account or give it to an agent. Most working sessions end when you close the tab, so download anything you want to keep. Audio Workspace can recover up to five original recordings and their edit settings from this browser's local storage; use Clear session to remove them.

The website measures which tool pages people open, category choices, starts, successes, broad error types, copies, and downloads. Search measurement uses only the length of the phrase and the number of results in broad ranges. Analytics do not include your search phrase, files, filenames, tool inputs, or results. Tool pages disable session replay, console recording, and automatic error capture.

Some tasks load code or model files from third parties. Image OCR, image tag suggestions and captions, small-image AI enhancement, speech transcription, PDF narration, and local text and chat models download model files to the browser on first use. The model then processes your chosen input locally. Image tag scores compare only the labels you entered; they are not independent probabilities. Image captions, enhanced details, and chat replies can be inaccurate. A large download or a device with insufficient memory can prevent an AI task from running. Browser voice playback depends on a local voice being available on the device; PDF narration instead uses a downloaded local voice model. Each tool page explains its own format, size, browser, and accuracy limits.

Peer calls, file sharing, and the whiteboard use WebRTC. Both people exchange invitation and answer codes manually and keep their tabs open. A public STUN service can see connection metadata, including network addresses; file contents and live strokes travel on the encrypted peer connection. Some networks cannot establish a direct connection without a TURN relay, which the free tools do not provide. Browser text workflows chain local tools only while the tab stays open; you can save their step list as a JSON file.

PDF Workspace shows page thumbnails from local PDFs and lets you reorder and rotate individual pages before exporting. PDF protection uses a password to encrypt a copy. A visible typed signature is only a visual mark; it is not a cryptographic signature. PDF structural inspection checks whether a document opens and reports basic properties. The separate signature checker tests signed-byte integrity for a supported detached CMS signature; it does not establish certificate trust, signer identity, revocation status, timestamps, or whether later changes are permitted. PDF text extraction into Word or Excel does not restore the original layout or table structure. PDF narration reads editable extracted text aloud and exports WAV; it does not summarize the document or make a conversation. Scanned PDFs need OCR first. Redaction lets you draw several areas across the page previews, then rasterizes every page to remove the original selectable text layer. Always inspect the downloaded result before sharing sensitive material.

Every tool page includes task-specific instructions, limits, and answers to common questions. The questions describe the controls and outputs available in that browser tool.

Choosing a tool

• Developer: format data, decode and optionally verify JWT signatures, generate QR codes and passwords, assess password strength, and run bounded code snippets in an isolated browser sandbox.
• PDF: merge, split, rotate, add text or a visual signature, watermark, number, inspect signed-byte integrity, encrypt or unlock files you have a password for, compress losslessly, extract text, narrate selectable text, render pages, redact selected areas, and run OCR.
• Image: convert, compress, resize, enlarge with interpolation or a local model for small images, remove a portrait background, blur detected faces, draft a caption with a local model, compare an image with your own candidate tags, sample colors, inspect selected metadata, create a PDF, and recognize English text locally.
• Audio: work on up to five recordings in recoverable local tabs, trim, convert to supported browser formats, change speed or pitch, reduce steady noise, compress, write timestamped notes, and transcribe up to two minutes of English speech locally.
• Writing and SEO: count and clean text, preview Markdown, draft metadata and sitemaps, and inspect HTML you provide. The SEO audit does not crawl a live website.
• Agents: review AGENTS.md structure, validate a common mcpServers JSON shape, estimate prompt size, explain five-field cron expressions, and chat with a small temporary local model.
• Collaboration: connect two browsers for a peer video call, direct file transfer, or a temporary shared whiteboard; chain local text tools in a browser workflow.

These tools are task-specific. Results from OCR, manual transcription notes, PDF extraction, compression, SEO checks, and generated code need your review. Model output and heuristic scores are not guarantees.

When you need an agent to work on files, use services, or keep running after you disconnect, explore Matrix OS and its quickstart.