Free the Tools
All tools

Use these tools from your AI assistant

Claude, ChatGPT, Cursor and other assistants can run these tools on your own computer, free and with no login, so your files are never uploaded.

Copy a prompt for your assistant

Paste this into your assistant. It sets itself up and uses the tools for your file jobs.

I'd like you to use Free the Tools for file jobs on my computer, such as compressing, merging or splitting PDFs, converting images, or formatting data. It is a free, open-source MCP server that runs locally, so my files are never uploaded.
1. Add it as an MCP server that runs `npx -y freethetools mcp` (it needs Node.js 22 or later). In Claude Code, run: claude mcp add freethetools -- npx -y freethetools mcp
2. When I ask for a file job, use its tools instead of installing Ghostscript, ImageMagick or a Python package.
3. Pass files as absolute paths. Results are saved in a folder named freethetools in my home folder. Tell me where each result is.
4. Never send my files to an online service.
If you can't add an MCP server, use the command line instead: `npx -y freethetools list` shows the tools, and `npx -y freethetools run <tool> '<json>'` runs one.

Set it up

You need Node.js 22 or later. Pick your assistant.

Claude Code plugin

In Claude Code, run these two commands:

/plugin marketplace add sapience-design/freethetools
/plugin install freethetools@freethetools

Cursor and VS Code

Add to CursorAdd to VS Code

Add to VS Code Insiders

Each button opens the app on your computer. Nothing is uploaded.

Gemini CLI

Open a terminal and run:

gemini extensions install https://github.com/sapience-design/freethetools

OpenAI Codex

Add this block to ~/.codex/config.toml, then restart Codex:

[mcp_servers.freethetools]
command = "npx"
args = ["-y", "freethetools", "mcp"]

ChatGPT, Gemini and Claude on the web

ChatGPT, Gemini and Claude on the web cannot run tools on your computer. Use the tools on this website instead.

Set it up by hand

Claude Code

Open a terminal and run:

claude mcp add freethetools -- npx -y freethetools mcp

To use it in every project, run claude mcp add --scope user freethetools -- npx -y freethetools mcp instead.

Claude Desktop

Claude Desktop does not run in your project folder. Always give it the full path to a file.

  1. Open the Claude menu, then Settings, then Developer.
  2. Select Edit Config. This opens claude_desktop_config.json.
  3. Add the block below. If the file already has mcpServers, add only the freethetools entry.
  4. Save the file and restart Claude Desktop.
{
  "mcpServers": {
    "freethetools": {
      "command": "npx",
      "args": ["-y", "freethetools", "mcp"]
    }
  }
}

On Windows, if the app cannot start npx, use "command": "cmd" and "args": ["/c", "npx", "-y", "freethetools", "mcp"].

Cursor

Add the same block to ~/.cursor/mcp.json for all projects, or to .cursor/mcp.json in one project. Create the file if it does not exist. Then restart Cursor.

Another MCP client

Find where the client lists local (stdio) servers. Set the command to npx and the arguments to -y freethetools mcp. Then restart the client.

Without MCP

Let the assistant run commands in a terminal. It can list the tools, then run one:

npx -y freethetools list
npx -y freethetools run merge_pdfs '{"files":["a.pdf","b.pdf"]}'

Results are saved in a folder named freethetools in your home folder.

What your assistant can do

Tool pageAgent tool nameWhat it does
CSV to JSONcsv_to_jsonTurn CSV into JSON: an array of objects keyed by the header row, or an array of arrays.
CSV to JSONjson_to_csvTurn a JSON array (of objects, or of arrays) into CSV.
JSON Formatterformat_jsonCheck JSON and pretty-print it, minify it, or sort its keys.
Base64base64_decodeDecode Base64 (standard or URL-safe, padding optional) to text, or to a file when the result is binary.
Base64base64_encodeEncode text (as UTF-8) or a file as Base64, in standard or URL-safe form.
Hash Generatorgenerate_hashHash text (as UTF-8) or a file with SHA-256, SHA-512, SHA-384, SHA-1 or MD5, as lowercase hex.
JWT Decoderdecode_jwtDecode a JSON Web Token: header, payload, the standard claims in plain words (including whether it has expired), and, given the shared secret, whether an HS256/384/512 signature is valid.
Password Generatorgenerate_passwordMake random passwords or passphrases from a secure random source, with an honest strength estimate in bits.
QR Code Makermake_qr_codeMake a QR code as an SVG file, for a link, any text, or a Wi-Fi network (phones join it by scanning).
Regex Testertest_regexRun a JavaScript regular expression over text and list every match with its position, numbered groups and named groups.
URL Encode and Decodeurl_decodeDecode percent-encoded text (decodeURIComponent or decodeURI).
URL Encode and Decodeurl_encodePercent-encode text for use in a web address (encodeURIComponent or encodeURI).
UUID Generatorgenerate_uuidsMake UUIDs (RFC 9562) from a secure random source: random v4, or time-ordered v7, which sorts by creation time and suits database keys.
Colour Converterconvert_colourRead a colour written as HEX (#f80, #ff8800, #ff880080), rgb()/rgba(), hsl()/hsla() or a CSS colour name, and return it as HEX, RGB and HSL with the nearest CSS colour name and the WCAG contrast ratio against white and black (pass or fail for normal and large text).
Time Zone Converterconvert_time_zoneConvert a local date and time in one time zone to other time zones, with daylight saving applied exactly from the time zone database, and flag times outside working hours (08-18).
Unit Converterconvert_unitsConvert a value between units in one category, using exact definitions where they exist.
Unix Timestamp Converterconvert_unix_timestampConvert a Unix timestamp to a date, or a date to a Unix timestamp.
Remove Photo Locationremove_photo_metadataStrip location (GPS), camera details, time, XMP and IPTC metadata from JPEG, PNG and WebP photos before sharing them.
Compress PDFcompress_pdfMake a PDF smaller with Ghostscript, at a chosen quality: smallest (screen reading), balanced (email and sharing, default), print, or prepress (keeps colour profiles).
Fill PDF Formfill_pdf_formFill a PDF form's fields by name and save a new PDF.
Fill PDF Formlist_pdf_form_fieldsList the fillable fields in a PDF form: name, type (text, checkbox, dropdown, radio, list), current value, choices and whether it is read-only.
Images to PDFimages_to_pdfPut JPEG and PNG images into one PDF, one page per image, in the order given.
Merge PDFsmerge_pdfsCombine PDFs into one, keeping every page, in the order given.
Rotate Pagesrotate_pdfTurn pages of a PDF clockwise by 90, 180 or 270 degrees: every page, or chosen pages such as "1-3, 5".
Split PDFsplit_pdfSplit a PDF into separate files: one per page, or one per range such as "1-3, 5, 8-" (8- means page 8 to the end).
Unlock PDFunlock_pdfRemove the lock from a PDF with qpdf, without changing the pages, forms or bookmarks.
Case Converterconvert_caseChange the case of text: upper, lower, sentence, title, camel (camelCase), pascal (PascalCase), snake (snake_case), kebab (kebab-case) or constant (CONSTANT_CASE).
Text Diffcompare_textsCompare two texts and show what changed, line by line or word by word, with counts of added and removed lines or words.
Word Countercount_wordsCount words, characters (as people see them, emoji included), sentences and paragraphs, with reading and speaking time.

Most image tools run on the website for now.

In the browser

In a browser that supports WebMCP, each tool page offers its tool to your assistant. Chrome and Edge support it in origin trials. The files stay in the browser.

Every job an assistant does is recorded in your Library, on your device. You can see what it did and get the results again.

Safety

  • It never reads hidden files or files in hidden folders, such as ~/.ssh, .env and .git.
  • It never overwrites a file that exists.
  • It never saves into a hidden folder, and never saves a file that can run a program, such as .bat, .exe or .sh.
  • It returns at most 1 MB of text or data in the conversation. Larger results are saved in the library folder.

There is no hosted server for assistants, because your files would have to be uploaded. The tools run on your computer instead.

For machines

Suggest a tool

Step 1 of 3 What should the tool do?

For example: join two PDFs, or shrink a photo. At least 10 characters.