Paste any X article, thread, or tweet. Get clean Markdown that ChatGPT and Custom GPTs read as plain prose — short enough for one message, structured enough to cite.
ChatGPT can browse URLs, but x.com serves a JavaScript-rendered page with a login wall to anyone who isn't signed in — including the OpenAI fetcher. The model often ends up summarising the page chrome ("Sign in to X", "Trending now") instead of the actual post. The cleanest workaround is to hand ChatGPT the text directly.
xmd resolves the post server-side and returns Markdown. ChatGPT does not need to browse at all — it just reads what you paste.
Convert the URL at xmdbot.com, copy the Markdown, paste into the chat. A typical thread fits inside a single message — far below the context window of any GPT-4-class model — so you get the whole post in one shot.
Custom GPTs accept up to 20 files of structured knowledge. Save xmd's output as.md and drop it into the GPT builder's knowledge panel. Each post lands as a single document with the URL and author in the front-matter, so when the GPT answers a question it can cite the original X post — useful for research GPTs built around a specific account, list, or topic.
Practical pattern: pin 10–15 of an account's best long-form X articles into a Custom GPT and you have a "talk to this thinker" assistant that quotes them accurately. xmd's clean Markdown is what makes the citations meaningful.
An HTTP API for the OpenAI Assistants API, function-calling tools, and bulk ingest into a vector store — converting X URLs in batch from a Python or Node client — is on the waitlist.
A single Markdown document: title, author handle, timestamp as front-matter; body as ordered sections; quoted tweets as nested blockquotes; no script tags, no tracking parameters, no HTML entities. ChatGPT treats it the same as any Markdown file you might upload from your laptop.
Paste the X URL into the xmd converter, copy the Markdown, and paste it into ChatGPT. The Markdown is short enough to fit comfortably inside a single message and clean enough that ChatGPT does not waste tokens parsing HTML.
Yes. ChatGPT Custom GPTs accept Markdown files in their knowledge configuration. xmd outputs one .md file per article or thread, with the post URL and author handle in the front-matter so the GPT can cite them back. Useful for building research GPTs around a specific account or topic.
In practice, no. ChatGPT can fetch x.com URLs, but X serves a JavaScript-rendered page with a login wall to anonymous clients. The model often summarises the page chrome instead of the post body. Converting through xmd first gives ChatGPT the actual content.
The web converter is shipped today. A programmatic HTTP API for OpenAI Assistants, function-calling tools, and bulk ingestion into vector stores is on the waitlist — join from this page to get notified when it ships.
Yes for the paste-into-conversation flow — it works on any plan because the Markdown is just text. Custom GPTs require a Plus, Team, or Enterprise plan to build, but anyone with a link can use a published GPT that has X-derived knowledge.
Ready to try it? Open the xmd converter and paste an X URL. The Markdown lands in your clipboard in under 200ms — paste it straight into ChatGPT.