Single posts, fully resolved
Likes, reposts, replies, quotes, bookmarks and views. Poll results with percentages. Quoted posts unnested. Alt text most scrapers throw away.
Read the endpointRead X without asking permission
Paste any post, thread or profile. Get back clean Markdown for your prompt, or strict JSON for your database. No account, no API key, and nothing to pay.
Works with x.com and twitter.com links, or a bare post id.
Try one
Whatever is doing the reading, it can reach this
Claude, ChatGPT, Cursor, your own script, a cron job, a Discord bot. Nothing needs an account here, so nothing needs to be told a secret. One HTTP request, five formats, no handshake.
Claude Code
Codex
Gemini CLI
Cursor
XCrap.CC
JSON
Markdown
CSV
No key is exchanged in either direction.
X priced everyone out, then locked the door
The official API starts at $200 a month and climbs to $42,000. Below that tier you get almost nothing. Meanwhile x.com stops a logged-out reader after two posts and asks them to sign in. Neither of those is a technical limit. They are a business decision, and it lands hardest on the people building small tools and research projects.
Every post XCrap reads is one anybody could read in a browser. We just do not make you log in to read it.
Three steps, no setup
-
01
Hand it a link
A post, a thread, a profile. Through the box above, the REST API, the MCP server, or a one-line curl.
-
02
We go and get it
Our client talks to X's own internal endpoints with rotating anonymous tokens. Three more sources sit behind it, so one of them breaking is not your problem.
-
03
Take the shape you want
Markdown for a prompt. JSON for a database. YAML, CSV or a standalone HTML card. Set it with one query parameter or an Accept header.
Everything the official API charges for
Threads that actually unroll
Paste the first post, not the last one. X only exposes a post's parent, so we stitch the author's timeline by conversation id to walk it forwards.
How unrolling works
Media you can keep
Direct MP4 links at every bitrate X encoded, and images at the resolution they were uploaded, not the resized preview. Streamed through, never stored.
Media endpoint
Fifty links at once
One request, up to fifty posts. A dead link returns an error for that entry and the other forty-nine still come back.
Bulk endpoint
Five formats, one parameter
Add ?format=markdown, or send an Accept header. Markdown responses carry an x-markdown-tokens header so an agent can budget its context before it reads.
- GET
/v1/tweet?url=…&format=markdown- content-type
text/markdown- x-markdown-tokens
49
MCP
Plugs straight into your agent
An MCP server for Claude Code, Claude Desktop and Cursor. Or download the skill file and teach any model the API in one paste.
# Claude Code — one line
claude mcp add xcrap -- npx -y @xcrapcc/mcp
# Claude Desktop, Cursor — config block
{
"mcpServers": {
"xcrap": {
"command": "npx",
"args": ["-y", "@xcrapcc/mcp"]
}
}
}
Ask for the shape you need
# jack (@jack)
**2006-03-21 20:50 UTC** · [permalink](https://x.com/jack/status/20)
just setting up my twttr
308,067 likes · 124,684 reposts · 18,017 replies{
"id": "20",
"text": "just setting up my twttr",
"author": { "screen_name": "jack" },
"metrics": { "likes": 308067 }
}id: '20'
text: just setting up my twttr
author:
screen_name: jack
metrics:
likes: 308067id,url,created_at,author_handle,text,likes
20,https://x.com/jack/status/20,2006-03-21,jack,"just setting up my twttr",308067<!-- a standalone card, no stylesheet required -->
<article class="xcrap-post">
<a href="https://x.com/jack/status/20">@jack</a>
<p>just setting up my twttr</p>
</article>Same post, same request, five shapes. The header does it too: Accept: text/markdown.
Four ways in. Pick one.
MCP server
Add one block to your config and your agent can read X on its own.
claude mcp add xcrap -- npx -y @xcrapcc/mcp
Skill file
A single Markdown file that teaches any model the whole API. No install.
curl xcrap.cc/skills.md
OpenAPI 3.1
A machine-readable spec every generator and agent framework already understands.
GET /openapi.json
SDKs
npm install @xcrapcc/sdk, or pip install xcrap-sdk. Typed, tiny, no dependencies.
npm i @xcrapcc/sdk · pip install xcrap-sdk
All four talk to the same endpoints. Pick whichever fits how you already work.
The ones people actually ask
What is XCrap?
A free service that reads public posts, threads and profiles from X and hands them back as Markdown, JSON, YAML, CSV or HTML. It exists so that agents, bots and small projects can read X without an API key.
Is it really free?
Yes. No account, no key, no card, no trial. It is paid for by donations, and the running costs are a domain and one small server.
Can it read private or protected accounts?
No, and it never will. Protected accounts return the same 404 they would give anyone else, and direct messages are out of reach entirely. XCrap reads public accounts only: their posts, profiles, replies and follower lists.
Can I download the video or images?
Yes. /v1/media lists every attached file with a download link, at every bitrate X encoded. Files stream straight through and are never written to our disk.
Is there full-text search?
Yes: /v1/search takes any query X search understands, operators included (from:, "exact phrase", since: and the rest), and returns the latest or top posts, or only those with photos or videos. It has the tightest rate limit of any endpoint.
Is this legal?
Reading public web pages is not unauthorised access, which US courts confirmed in hiQ v. LinkedIn. It does break X's terms of service, which is a contract matter between us and X, not a criminal one. You are responsible for what you do with the data you receive.
Are you featured anywhere?
Aura++ lists XCrap in its directory, and its badge is below. If you run a directory and want to add it, write to [email protected] and your badge goes up beside that one.
It is free and there is nothing to sign up for
No key to generate, no card to add, no email to confirm. Paste a link and see what comes back.
