Let your AI apps use Toolbox
What it does
AI apps are good at understanding what you want, but they cannot compress a video or put a password on a PDF by themselves. Toolbox 2.0.0 gives them 36 of its tools through MCP, the standard way AI apps connect to other software. You ask in plain words, the AI picks the tool and the settings, and Toolbox does the work on your Mac.
Things you can ask:
- “Compress the videos on my Desktop so I can email them.”
- “Merge these three PDFs into one and add a password.”
- “Transcribe this interview and save it as subtitles.”
- “Download this video as an MP3.”
- “Remove the background from every photo in this folder.”
- “Make this scanned contract searchable, then turn it into a Word file.”
The AI app gets back the new file names, how much smaller they are, or the text it asked for. A tool call starts in about 10 ms, and Toolbox runs two heavy jobs at a time, so your Mac stays responsive while a batch runs.
Set it up in 3 steps
- Open Toolbox and go to Settings › Agents. Make sure Allow AI apps to use Toolbox is on.
- Click Connect next to the AI app you use.
- Restart that app. Then ask it to do something with your files.
Toolbox needs to be activated with your license. If you do not have one yet, see pricing.
Which AI apps work
One-click Connect: Claude, Claude Code, Cursor, VS Code (Copilot), Windsurf, Codex, LM Studio.
ChatGPT can’t use Toolbox yet. It only connects to online MCP servers, and Toolbox’s tools run on your Mac.
Any other app that supports MCP works too. Click Copy Settings in Settings › Agents, or paste this into the app’s MCP settings:
{
"mcpServers": {
"toolbox": {
"command": "/Applications/Toolbox.app/Contents/MacOS/Toolbox",
"args": ["mcp"]
}
}
}The 36 tools
Each one runs the same engine as the tool in the Toolbox window, so results match the app. Tools that need you to point or draw, such as signing, redacting or recording, stay in the app.
Images (7)
| Tool | What it does |
|---|---|
Convert Imagesconvert_images | Converts images to JPEG, PNG, HEIC, TIFF or WebP. Reads almost any image format, including RAW, HEIC, WebP, GIF, BMP and PSD. |
Compress Imagescompress_images | Makes photos and images smaller without visible loss. If a file can't be made smaller, the original is copied unchanged. |
Resize Imagesresize_images | Resizes images by percentage, by their longest side, or to an exact width and height. Keeps the format. |
Rotate Imagesrotate_images | Rotates images in 90° steps and/or flips them, without quality loss where the format allows. |
Remove Backgroundremove_image_background | Cuts out the main subject of each photo (people, pets, objects) and saves it as a PNG with a transparent, white or colored background. |
Remove Metadataremove_image_metadata | Removes location, camera, date and other hidden metadata from images before sharing them. Pixels are not re-encoded. |
Watermarkadd_watermark | Adds a text watermark, like your name or “© 2026 Studio”, to photos, PDFs and videos, in a corner or tiled across. Photos keep their format and metadata, PDFs keep their text. Videos need FFmpeg. |
PDF (11)
| Tool | What it does |
|---|---|
Merge PDFsmerge_pdfs | Combines PDFs (and images, which become pages) into one PDF, in the order given. |
Split PDFsplit_pdf | Splits a PDF into single pages, into files of N pages each, or into the page ranges you list. |
Make PDF Fillablemake_pdf_fillable | Turns flat PDF forms into fillable ones: finds the blanks (lines to write on, empty boxes and checkbox squares) on every page and adds real form fields there, named after their labels. |
Compress PDFcompress_pdf | Makes PDFs smaller by recompressing their images. Text stays selectable. If a PDF can't be made smaller, the original is kept. |
Add Passwordadd_pdf_password | Encrypts PDFs so they need a password to open. Can also block printing or copying text. |
Remove Passwordremove_pdf_password | Saves unlocked copies of password-protected PDFs. Needs the current password. |
PDF to Imagespdf_to_images | Saves every page of a PDF as an image. PDFs with several pages get a folder with one image per page. |
Images to PDFimages_to_pdf | Turns images into one PDF, one image per page, in the order given. |
Make PDF Searchablemake_pdf_searchable | Adds an invisible text layer to scanned PDFs, so they can be searched and copied. Pages keep their look. |
PDF to Wordpdf_to_word | Converts PDFs (text or scanned) into editable Word, RTF or plain-text documents. |
Markdown to PDFmarkdown_to_pdf | Typesets Markdown files as print-ready PDFs, with headings, lists, tables, code and images. |
Text (OCR and transcription) (2)
| Tool | What it does |
|---|---|
Extract Text (OCR)extract_text | Reads the text in images, screenshots, photos of documents and PDFs, including scans. Long results are also saved as .txt files. |
Transcribetranscribe | Turns speech in an audio or video file into text. Can also save it as .txt or .srt subtitles. |
Video (11)
| Tool | What it does |
|---|---|
Compress Videocompress_video | Makes videos much smaller for sharing, email or upload, with optional downscaling. Uses FFmpeg. |
Convert Videoconvert_video | Converts videos to MP4, MOV, WebM or M4V. Copies streams without re-encoding when possible. Uses FFmpeg. |
Resize Videoresize_video | Resizes videos to a preset (4K, 1080p, 720p, 480p, square, vertical 9:16) or a custom size. Uses FFmpeg. |
Trim Videotrim_video | Keeps only the part of a video between a start and end time, without re-encoding when possible. |
Change Speedchange_video_speed | Speeds videos up or slows them down (0.25× to 4×), keeping the voice pitch natural. Uses FFmpeg. |
Video to GIFvideo_to_gif | Turns a video, or part of one, into a good-looking animated GIF. Uses FFmpeg. |
Mute Videomute_video | Removes the sound from videos without re-encoding the picture. |
Merge Videosmerge_videos | Joins videos one after another into a single MP4, in the order given. |
Extract Framesextract_frames | Saves still images from videos: one every N seconds, a number spread evenly, or the single frame at a given time. |
Download Videodownload_video | Downloads videos or just their audio from YouTube, X, Instagram, TikTok, Vimeo, Reddit and 1,000+ other sites, and can pick the audio language on videos with dubbed tracks. Uses yt-dlp. |
Media Infomedia_info | Reports duration, resolution, frame rate, codecs, bitrate, HDR, audio channels, tracks and metadata of video and audio files. |
Audio (4)
| Tool | What it does |
|---|---|
Convert Audioconvert_audio | Converts audio files to MP3, M4A (AAC), WAV, FLAC or AIFF. Uses FFmpeg. |
Extract Audioextract_audio | Saves the sound track of videos as audio files, without quality loss when possible. Uses FFmpeg. |
Clean Up Audioclean_up_audio | Removes background noise, hum and hiss from voice recordings, evens out loudness to −16 LUFS and can cut long silences. Uses FFmpeg. |
Trim Audiotrim_audio | Keeps only the part of an audio file between a start and end time, with optional fade in and out. |
Status (1)
| Tool | What it does |
|---|---|
Toolbox Statustoolbox_status | Shows the Toolbox version, whether it is activated, whether FFmpeg and yt-dlp are installed, and where results are saved. AI apps check this when a tool reports something missing. |
From the command line
The same tools work in Terminal and in scripts. In Settings › Agents, click Install toolbox command. Then:
toolbox compress-images ~/Desktop/*.jpg --quality high toolbox merge-pdfs a.pdf b.pdf --name Report toolbox extract-text scan.pdf toolbox download-video "<link>" --quality 1080 toolbox list toolbox help compress-video
toolbox list shows every tool, toolbox help <tool> shows its options, and --json prints results as JSON for scripts.
Your files stay on your Mac
- Every tool runs on your Mac. Toolbox never uploads your files.
- Your originals are never changed. Results go to a dated folder in Downloads, unless the AI app picks another folder.
- Video and audio tools use FFmpeg, which is free and installs in one click from Toolbox. Download Video uses yt-dlp, which Toolbox installs by itself.
- You choose which tools AI apps can use: in Settings › Agents, each tool has its own switch, plus Turn All On and Turn All Off. Changes apply straight away, without restarting the AI app.
- Turn off Allow AI apps to use Toolbox at any time and no AI app can run a tool.
More on what the app connects to is on the security page.
Questions
Is my data sent anywhere?
Toolbox does the work on your Mac and never uploads your files. It sends the AI app a short result: the new file names and sizes, and the text itself for Extract Text and Transcribe. Your AI app may send that result to its own servers, like anything else you type into it. For fully offline use, pair Toolbox with a local model in LM Studio.
Does it need a license?
Yes. AI apps can use Toolbox once Toolbox is activated on that Mac. Every license includes it, with no extra cost.
Which AI apps work?
Claude, Claude Code, Cursor, VS Code with Copilot, Windsurf, Codex and LM Studio have a one-click Connect button. Any other app that supports MCP works too: click Copy Settings and paste the JSON into that app's MCP settings. ChatGPT can't use Toolbox yet, because it only connects to online MCP servers and Toolbox's tools run on your Mac.
Does it need FFmpeg?
Only the video and audio tools do. FFmpeg is free, and Toolbox installs it in one click. Download Video uses yt-dlp, which Toolbox installs by itself. Image, PDF and text tools need nothing extra.
Will it change my original files?
No. Toolbox only writes new files. Results go to a dated folder in Downloads, unless the AI app picks another folder.
Can I turn it off?
Yes. Switch off "Allow AI apps to use Toolbox" in Settings › Agents and no AI app can run a tool. You can also turn single tools on or off there, so AI apps only see the ones you want.
102 Mac tools. One app. Pay once.
From $19 one-time
Buy Toolbox from $19macOS 15 Sequoia or later · Notarized by Apple · 7-day refund