x-reader

Read X/Twitter posts and replies in Claude Code — no Twitter API key needed.

How It Works

Step	Source	What You Get
Tweet content	ADHX public API	Full text, author, engagement stats
X Articles	ADHX article field	Complete Markdown with images
Replies	Chrome DevTools network interception	~30 top replies per page load

Tweet content is fetched via the free ADHX API (one curl, no auth).

Replies are captured by intercepting X.com's own TweetDetail GraphQL request through Chrome DevTools MCP — the same data X's frontend uses to render the reply section. This uses your logged-in Chrome session, so you see everything your account can see.

Install

Claude Code (one-liner)

mkdir -p ~/.claude/skills/x-reader && \
  curl -sL https://raw.githubusercontent.com/XHXIAIEIN/x-reader/main/skills/x-reader/SKILL.md \
  -o ~/.claude/skills/x-reader/SKILL.md && \
  mkdir -p ~/.claude/skills/x-reader/scripts && \
  curl -sL https://raw.githubusercontent.com/XHXIAIEIN/x-reader/main/skills/x-reader/scripts/parse_replies.py \
  -o ~/.claude/skills/x-reader/scripts/parse_replies.py

Other agents

Drop the skills/x-reader/ folder into your agent's skills directory:

Agent	Skills directory
Claude Code	~/.claude/skills/
Cursor	.cursor/skills/ or ~/.cursor/skills/
Gemini CLI	~/.gemini/skills/
GitHub Copilot	.github/skills/

Usage

Paste any X/Twitter link into Claude Code:

https://x.com/karpathy/status/2015883857489522876

Or invoke directly:

/x-reader https://x.com/karpathy/status/2015883857489522876

Claude will automatically fetch the tweet content and replies.

Supported URL formats

• https://x.com/{user}/status/{id}
• https://twitter.com/{user}/status/{id}
• https://x.com/{user}/article/{id}

Prerequisites

For tweet content only

None. The ADHX API is free and requires no authentication.

For replies (optional)

• Chrome DevTools MCP plugin installed in Claude Code
• Chrome browser open with DevTools debugging enabled
• Logged into X.com in Chrome

Without Chrome DevTools, the skill gracefully degrades to tweet-content-only mode.

How reply interception works

Browser loads tweet page
        |
        v
X.com frontend requests TweetDetail GraphQL API
        |
        v
Chrome DevTools MCP captures the network response
        |
        v
parse_replies.py extracts structured reply data
        |
        v
Claude presents replies with author + text + likes

Each page load returns approximately 30 top replies ranked by relevance. No scrolling or pagination needed — the data comes directly from X's API response.

File structure

skills/x-reader/
├── SKILL.md                  # Skill instructions
└── scripts/
    └── parse_replies.py      # TweetDetail GraphQL response parser

Limitations

• Reply interception requires Chrome DevTools MCP (not available in all environments)
• Each page load returns ~30 replies (X's default batch size)
• ADHX is a third-party service — if it goes down, the skill falls back to fxtwitter API
• Article images are remote URLs, not downloaded locally

License

MIT