Document Tools in ChatGPT

Questions it answers

Four small document jobs in one place: count the words in a text, translate a web page you own into another language, make a calendar file, or turn a table into a CSV file.

Ask "how many words and characters is this?", "will this fit in one SMS?", "translate my docs page into Vietnamese, I own it", "make a calendar file for a meeting every Monday at 10 in New York" or "put these rows in a CSV I can open in Excel".

Counting gives characters, words, sentences, paragraphs, lines, SMS segments, reading time and Flesch reading scores, and states the rule behind each count.

Calendar files follow RFC 5545 and open in Google Calendar, Outlook and Apple Calendar, with time zones, repeats, reminders and all-day events. CSV files follow RFC 4180, with correct quoting and an optional byte order mark for Excel.

Translation reads a public web page that you own or have the rights to translate, such as your documentation or public-domain text, and returns Markdown in about 100 languages with headings, lists and code blocks in place.

It does not add events to your calendar, open or edit an existing spreadsheet, or translate copyrighted fiction or comics. Files come back as text and as a download link that expires after 24 hours. Your text and pages are not stored.

What it can do

• Counts characters, words, sentences, paragraphs and lines, with the counting rules stated
• Reports SMS encoding and how many segments a message takes
• Estimates reading time and Flesch reading scores for English text
• Makes .ics calendar files with time zones, repeats, reminders and all-day events
• Makes CSV files with correct quoting and an optional Excel byte order mark
• Returns the file text and a download link that works for 24 hours
• Translates owned web pages, documentation, and public-domain text into clean Markdown without persistent copies

Tools

• count_text, Count words and characters. Use this when the user asks how long a text is, for example "how many words is this", "character count for this tweet", "will this fit in one SMS", "how long does this take to read". Input tips: pass text exactly as written, up to 20,000 characters (about 3,500 words); do not trim or fix it. Returns: characters, characters without spaces, graphemes (user-perceived characters), words, sentences, paragraphs, lines, UTF-8 bytes, SMS encoding, units and segments, reading minutes, and Flesch reading ease and grade (English only, null otherwise), with the rule used for each count. Not for: counting text inside a file, PDF or spreadsheet the tool cannot see (paste the text instead), or counting tokens for a model. Errors: a text over the limit is refused; split it and add the counts. The text is not stored.
• make_calendar_file, Make a calendar (.ics) file. Use this when the user wants a calendar file to import, for example "make a calendar file for this meeting", "create an .ics for my weekly class", "export these dates so I can import them". Input tips: pass 1 to 50 events, each with a title, a start (2026-11-02 for all day, or 2026-11-02T09:30) and an end or duration_minutes, plus timezone (IANA name such as Europe/Paris) for a time without Z or an offset; ask the user for the time zone, never guess it. Example event: {"title":"Sprint review","start":"2026-11-02T09:30","duration_minutes":45,"timezone":"Europe/Paris"}. Optional: location, description, reminder_minutes and recurrence (freq daily, weekly, monthly or yearly, with interval, until or count, and byday). Returns: the .ics text, a download link that works for 24 hours, its expiry and the event count. Not for: adding, changing or reading events in Google Calendar, Outlook or any calendar account; it only makes a file. Errors: status invalid with a message naming the event and field to fix.
• make_csv_file, Make a CSV file. Use this when the user wants a new CSV file from a table, for example "put this table in a CSV", "make a spreadsheet file from these rows", "export this list as CSV for Excel". Input tips: pass columns (header names) and rows, where each row has exactly one cell per column (text, number, true, false or null for empty). Example: columns ["name","city"], rows [["Ada","London"],["Linus","Helsinki"]]. Up to 2,000 rows and 100 columns. Optional: filename, excel_bom (true when the user opens it in Excel and the text has accents) and escape_formulas (show cells starting with = + - or @ as text). Returns: the CSV text, a download link that works for 24 hours, its expiry, the row count and byte size. Not for: opening, reading, editing or adding up an existing spreadsheet or .xlsx file; it only writes a new CSV. Errors: status invalid with a message naming the row and cell to fix.
• translate_page, Translate a webpage or documentation. Use this when the user wants to translate a public webpage, documentation or public-domain text that they own or have the rights to translate. Input tips: url of the page (http or https), target_lang as a language code such as vi, es, fr, de, ja, zh, pt, ko, and confirm_rights true only after the user confirms the rights. Returns: clean Markdown with headings, lists and code kept, usually in 5 to 15 seconds for a page of up to 8,000 characters, plus characters left today. Not for: copyrighted fiction, manga, manhwa or commercial comics, pasted text, local files or pages behind a login. Errors and next step: rights_not_confirmed (ask the user, then set confirm_rights true); page_too_large over 12,000 characters (offer a shorter page, such as one chapter on its own URL); fetch_failed or no_content (check the URL opens without a login, or pick a page whose text is in the HTML); daily_translation_spent (the daily character limit is used up; try again after 00:00 UTC); translation_failed (try again in a minute).

Add Document Tools in ChatGPT

ChatGPT does not read a local config file. A custom MCP server is added in the product, and only after developer mode is on. These steps were checked against OpenAI's plugin connect guide and the ChatGPT MCP guide on 2026-10-01. Developer mode can be hidden by a workspace policy, so a Team or Enterprise admin may have to allow it first.

1. Open ChatGPT settings, then Security and login, and turn on Developer mode.
2. Open ChatGPT plugins and select the plus button.
3. Enter a name such as "Document Tools" and a one-line description of what you want it to answer.
4. Under Connection, choose the public endpoint and paste this URL. Include the /mcp path. Do not add a tunnel id: this server is already public.

   https://docs.openkrill.app/mcp

5. Create the connection and review the tools ChatGPT lists. You should see count_text, make_calendar_file, make_csv_file, translate_page. The feedback tools on the same server are optional and not required for the lookup.
6. Start a new conversation, add the connection from the tools menu, and ask: How many words and characters is "The quick brown fox jumps over the lazy dog."?

No API key and no OAuth client id are required. The server answers without a login. Calls are limited per network and per day, and a limit comes back as a tool error rather than a partial answer. If the tool list is stale after a server change, open the connection and select Refresh, then start a new conversation.

The ChatGPT desktop app, Codex CLI and IDE extension share a different config. In the desktop app, open Settings, then MCP servers, choose Add server, select Streamable HTTP, and paste the same URL. In ~/.codex/config.toml the equivalent is a table named mcp_servers.docs with url = "https://docs.openkrill.app/mcp". ChatGPT on the web does not read that file. A published directory listing is also a different path: this page is the developer-mode connection, not a claim that Document Tools is already in the ChatGPT directory.

Treat the server as untrusted until you have read a few answers. A remote MCP server sees the arguments the model sends. Document Tools is read-only and documents what it sends onward on the privacy page.

If it does not work

A call checked on 2026-10-17

A short sentence is enough to show word and character counts. The text is not stored. The request below was posted to https://docs.openkrill.app/mcp as tools/call. HTTP 200. Source: the Document Tools server, read 2026-10-17.

{
  "method": "tools/call",
  "params": {
    "name": "count_text",
    "arguments": {
      "text": "Agent tools answer public questions."
    }
  }
}

The server returned structured content for this call.

Repeat the call yourself if you need a newer reading. Cached answers expire. A rate limit is not a result: wait and try again. Nothing in the call is a ranking, a filing, or advice.

Using it in production

Nothing to install and nothing to run: no package, no Python or Node.js version and no environment variable. Document Tools runs on our servers, and your client only needs the URL https://docs.openkrill.app/mcp.

Use it in other apps

• Document Tools in Claude
• Document Tools in Cursor
• Document Tools in VS Code
• Document Tools in Pi

Roles that use Document Tools

• Writer guide

Frequently asked questions

Words on this page

• MCP (Model Context Protocol) definition
• Streamable HTTP transport definition
• Tool primitive definition
• Rate limit policy definition

Source: plugin listing for docs in this repository, and a live tools/call on 2026-10-17. As of 2026-10-17.