Observed since 20 Oct, 23:18 UTC
No superseded metadata states collected. Unchanged observations do not create new revisions.
Give your AI Agent eyes to see the entire internet. Search + Read 10+ platforms.
Give your AI Agent instant access to the internet
Agent Search gives your AI agent the ability to search and read content from 10+ platforms — Twitter/X, Reddit, YouTube, GitHub, Bilibili, LinkedIn, web pages, and more — with a single install command.
The problem it solves:
| Pain Point | Reality |
|---|---|
| Twitter/X API | Paid, $215/mo for moderate use |
| Server IPs get 403'd | |
| YouTube transcripts | No official API |
| GitHub search | Auth required, rate limited |
| Web scraping | Returns raw HTML, not readable content |
Agent Search handles all of this transparently. Your agent calls a function, gets back clean readable text.
Paste this into your AI agent (Claude Code, Cursor, OpenAI Assistants, etc.):
Install Agent Search: https://raw.githubusercontent.com/travis-burmaster/agent-search-tool/main/docs/install.md
pip install agent-search-tool
| Platform | Search | Read | Auth Required |
|---|---|---|---|
| Web | ✅ | ✅ | No |
| GitHub | ✅ | ✅ | Optional |
| ✅ | ✅ | No | |
| YouTube | ❌ | ✅ (transcript) | No |
| Twitter/X | ✅ | ✅ | Cookie |
| ❌ | ✅ | Cookie | |
| Bilibili | ✅ | ✅ | Cookie |
| Exa Search | ✅ | ✅ | API key |
| RSS | ❌ | ✅ | No |
| V2EX | ✅ | ✅ | No |
# Check what's working
agent-search doctor
# Search the web
agent-search search "latest LLM benchmarks 2026"
# Configure optional channels
agent-search configure
Twitter search requires browser cookies. Two ways to set it up:
Option 1 — Auto-extract from your local browser (fastest):
agent-search configure --from-browser chrome
Automatically pulls Twitter cookies from Chrome. Also works with Firefox, Edge, Brave.
Option 2 — Manual via Cookie-Editor:
agent-search configure twitter-cookies "PASTE_COOKIE_STRING_HERE"
Once configured:
twitter search "AI agents 2026" -n 10
twitter read https://x.com/username/status/123456789
LinkedIn uses cookie-based auth via browser automation:
Option 1 — Auto-extract:
agent-search configure --from-browser chrome
Option 2 — Manual via Cookie-Editor:
agent-search configure linkedin-cookies "PASTE_COOKIE_STRING_HERE"
| Platform | Command |
|---|---|
Works out of the box — rdt search "query" | |
| YouTube | Works out of the box — yt-dlp --dump-json URL |
| GitHub | Works out of the box — gh repo view owner/repo |
| Web | Works out of the box — curl https://r.jina.ai/URL |
| Exa Search | agent-search configure exa-key YOUR_KEY |
| Bilibili | agent-search configure --from-browser chrome |
| Xiaoyuzhou Podcast | agent-search configure groq-key gsk_... |
agent-search doctor
Shows which channels are ready and exactly what to run to fix anything that isn't.
No hardcoded credentials. All cookies and API keys are stored in ~/.agent-search/config.yaml with 0600 permissions (owner read/write only). Credentials never leave your machine.
MIT — Copyright (c) 2026 Travis Burmaster
Download counts and popularity will appear here as the collector receives them.
| Version | Published | Direct dependencies | Open version |
|---|---|---|---|
| 1.4.8 | 2 Apr 2026 | 15 |
This describes our collected data; it is not a security assessment.
Shared repository data · GitHub
Observed since 20 Oct, 23:18 UTC
No superseded metadata states collected. Unchanged observations do not create new revisions.
package:e83dfc3b21f5d9c3acc7158ca697190fc606f9fa8e143607e11ae2c67d469123