fetch-threads-post

Extract all content from a Threads post using agent-browser (browser automation via CDP).

Supported URL Formats

• https://www.threads.net/@{username}/post/{shortcode}/
• https://www.threads.com/@{username}/post/{shortcode}
• https://www.threads.net/t/{shortcode}/

URLs may include query parameters (e.g. ?xmt=...). These are harmless and can be kept as-is.

Prerequisites

• agent-browser must be installed and available in PATH.
• If not installed: npm i -g agent-browser && agent-browser install

Procedure

Follow these steps in order. Use a dedicated session to avoid polluting other browser state.

Step 1 — Navigate to the Post

agent-browser --session threads open "<THREADS_URL>"
agent-browser --session threads wait --load networkidle
agent-browser --session threads wait 2000

The extra 2-second wait ensures dynamic SPA content finishes hydrating.

Step 2 — Check for Login Wall

agent-browser --session threads snapshot -i

Inspect the snapshot output:

• If you see a login form, "Log in" button, or modal overlay blocking content, the post requires authentication. Jump to Authenticated Access.
• If the post content is visible (you can see post text, author info), continue to Step 3.

Step 3 — Extract Post Author

agent-browser --session threads snapshot -i

From the snapshot, identify the author elements:

• Handle: Look for text starting with @ near the top of the post. Use agent-browser --session threads get text @ref on the element.
• Display name: Usually adjacent to or above the handle.

Step 4 — Expand Long Posts ("Read more")

Threads truncates long posts with a "Read more" link. You MUST check for this to avoid capturing incomplete content:

# Look for "Read more" in the snapshot
agent-browser --session threads snapshot -i

# If found, click it by text or ref
agent-browser --session threads click "Read more"
agent-browser --session threads wait 1500

Crucial: If the post remains truncated after clicking or the button is unreachable, use the get text body fallback in Step 5 to capture the hidden text from the DOM.

Step 5 — Extract Post Text

Identify the main post text container from the snapshot refs, then:

agent-browser --session threads get text @<post-body-ref>

If you cannot reliably identify the post container from the snapshot, or if the post is very long and spans multiple containers:

# Extract full page text to ensure no segment is missed
agent-browser --session threads get text body

Then parse the post content from the output. Verify that the extracted text matches the expected flow of the post (e.g., it doesn't end abruptly or skip sections).

Step 6 — Extract Media

From the snapshot, look for [img] or [video] elements within the post container:

# For images
agent-browser --session threads get attr @<img-ref> src

# For videos
agent-browser --session threads get attr @<video-ref> src

Collect all media URLs. If no media elements are found, note "No media" in the output.

Step 7 — Extract Engagement Metrics

From the snapshot, look for elements containing like/reply/repost counts. These are typically near the bottom of the post. Use get text on the relevant refs:

agent-browser --session threads get text @<likes-ref>
agent-browser --session threads get text @<replies-ref>
agent-browser --session threads get text @<reposts-ref>

Threads may display metrics as "123 likes" or just "123" with an icon. Extract whatever is visible.

Step 8 — Extract Timestamp

Look for a time element or relative date text (e.g. "2h", "Apr 10") in the snapshot:

agent-browser --session threads get text @<time-ref>
# Or try getting the datetime attribute for precise timestamps
agent-browser --session threads get attr @<time-ref> datetime

Step 9 — Extract Replies (if present)

Scroll down to load replies:

agent-browser --session threads scroll down 800
agent-browser --session threads wait 1500
agent-browser --session threads snapshot -i

For each visible reply, extract:

• Reply author handle
• Reply text content

Repeat the scroll → wait → snapshot → extract cycle until:

• No new replies appear in successive snapshots, OR
• You have collected a reasonable number of replies (20+ is sufficient unless the user asks for more)

# For each reply
agent-browser --session threads get text @<reply-author-ref>
agent-browser --session threads get text @<reply-text-ref>

Step 10 — Cleanup

agent-browser --session threads close

Step 11 — Assemble Output

Format the collected data as follows:

# Threads Post by @{handle}

## Author
- **Handle**: @{username}
- **Display Name**: {display_name}

## Post Content
{full_text_of_the_post}

## Media
- {media_url_1}
- {media_url_2}
(or "No media attached")

## Engagement
- Likes: {N}
- Replies: {N}
- Reposts: {N}

## Timestamp
{date_time}

## Source URL
{original_url}

---

## Replies

### @{reply_user_1}
{reply_text_1}

### @{reply_user_2}
{reply_text_2}

(... more replies ...)

Authenticated Access

If the post is behind a login wall:

1. Check for saved state:

   # If you have a previously saved Threads auth state, load it
   agent-browser --session threads state load threads-auth.json
   agent-browser --session threads reload
   agent-browser --session threads wait --load networkidle
   agent-browser --session threads wait 2000

1. If no saved state exists, inform the user:

"This Threads post requires login. Please run agent-browser --session threads --headed open <URL> to log in manually, then run agent-browser --session threads state save threads-auth.json to save the session for future use."

1. After auth state is loaded, return to Step 3 and continue extraction.

Troubleshooting

Content not loading

# Try a longer wait
agent-browser --session threads wait 5000
agent-browser --session threads snapshot -i

Obfuscated DOM / can't identify elements

Threads uses obfuscated class names. Always rely on the accessibility tree snapshot (snapshot -i) rather than CSS selectors. The snapshot provides semantic element types and visible text, which are stable even when class names change.

If the snapshot is too large or cluttered:

# Scope to main content area
agent-browser --session threads snapshot -d 4

CAPTCHA or rate limiting

If you encounter a CAPTCHA:

agent-browser --session threads screenshot captcha.png

Report to the user: "Threads is showing a CAPTCHA. Please solve it manually using agent-browser --session threads --headed open <URL>."

Fallback: Raw text extraction

If structured extraction fails, extract everything as raw text:

agent-browser --session threads get text body > /tmp/threads-raw.txt

Then parse the output manually to identify post content, author, and replies from the text flow.

Tips

• Session isolation: Always use --session threads to avoid interfering with other browser sessions.
• Snapshot is your eyes: Re-snapshot after every navigation, scroll, or wait. Refs are invalidated on page changes.
• Be patient with SPA: Threads is a heavy SPA. After networkidle, wait 1-2 extra seconds.
• Anti-Truncation: Long posts on Threads are notoriously difficult to capture via simple ref-based extraction. Always prefer get text body for long posts and verify the output against the visible post structure (using screenshot or snapshot) to ensure you haven't missed the middle or end of the text.
• Media quality: Image src attributes may point to CDN URLs with size parameters. These URLs are directly usable.