🇺🇸English

Reddit Fetch

Method 1: Gemini CLI (Try First)

Use Gemini CLI via tmux. It can browse, summarize, and answer complex questions about Reddit content.

Pick a unique session name (e.g., gemini_abc123) and use it consistently throughout.

Setup

tmux new-session -d -s <session_name> -x 200 -y 50
tmux send-keys -t <session_name> 'gemini -m gemini-3-pro-preview' Enter
sleep 3  # wait for Gemini CLI to load

Send query and capture output

tmux send-keys -t <session_name> 'Your Reddit query here' Enter
sleep 30  # wait for response (adjust as needed, up to 90s for complex searches)
tmux capture-pane -t <session_name> -p -S -500  # capture output

If the captured output shows an API error (e.g., quota exceeded, model unavailable), kill the session and retry without the -m flag (just gemini with no model argument). This falls back to the default model.

How to tell if Enter was sent

Look for YOUR QUERY TEXT specifically. Is it inside or outside the bordered box?

Enter NOT sent - your query is INSIDE the box:

╭─────────────────────────────────────╮
│ > Your actual query text here       │
╰─────────────────────────────────────╯

Enter WAS sent - your query is OUTSIDE the box, followed by activity:

> Your actual query text here

⠋ Our hamsters are working... (processing)

╭────────────────────────────────────────────╮
│ >   Type your message or @path/to/file     │
╰────────────────────────────────────────────╯

Note: The empty prompt Type your message or @path/to/file always appears in the box - that's normal. What matters is whether YOUR query text is inside or outside the box.

If your query is inside the box, run tmux send-keys -t <session_name> Enter to submit.

Cleanup when done

tmux kill-session -t <session_name>

If Gemini fails completely

If retrying without -m also fails, fall back to Method 2 below.

Method 2: curl with Reddit JSON API (Fallback)

Reddit's public JSON API works by appending .json to any Reddit URL. Use this when Gemini is unavailable (quota exhausted, API errors, etc.).

Listing hot/new/top posts

curl -s -H "User-Agent: Mozilla/5.0 (compatible; bot)" \
  "https://www.reddit.com/r/SUBREDDIT/hot.json?limit=15"

Replace hot with new, top, or rising as needed. For top, add &t=day (or week, month, year, all).

Fetching a specific post + comments

curl -s -H "User-Agent: Mozilla/5.0 (compatible; bot)" \
  "https://www.reddit.com/r/SUBREDDIT/comments/POST_ID.json?limit=20"

The response is a JSON array: [0] is the post, [1] is the comment tree.

Searching within a subreddit

curl -s -H "User-Agent: Mozilla/5.0 (compatible; bot)" \
  "https://www.reddit.com/r/SUBREDDIT/search.json?q=QUERY&restrict_sr=on&sort=new&limit=15"

Parsing the JSON

Use python3 inline to extract what you need:

curl -s -H "User-Agent: Mozilla/5.0 (compatible; bot)" \
  "https://www.reddit.com/r/SUBREDDIT/hot.json?limit=15" | python3 -c "
import json, sys
data = json.loads(sys.stdin.read())
for i, post in enumerate(data['data']['children'], 1):
    p = post['data']
    flair = p.get('link_flair_text', '') or ''
    print(f'{i}. [{flair}] {p[\"title\"]}')
    print(f'   {p[\"score\"]} pts | {p[\"num_comments\"]} comments | u/{p[\"author\"]}')
    print()
"

If curl returns empty or blocked

If the JSON API returns empty responses or HTTP 429, you may be rate-limited. Wait a moment and retry, or try with a different User-Agent string.

Weekly Installs

181

Repository

ykdojo/claude-code-tips

GitHub Stars

6.7K

First Seen

Jan 24, 2026

Security Audits

Gen Agent Trust HubPassSocketPassSnykWarn

Installed on

codex156

gemini-cli155

opencode154

cursor154

cline152

github-copilot151