Documentation Index
Fetch the complete documentation index at: https://docs.autosearch.dev/llms.txt
Use this file to discover all available pages before exploring further.
1. Install
npx autosearch-ai # one-shot install + init
# OR, if you already have the Python CLI on PATH:
autosearch init
init is an explicit step, not an npm postinstall hook. AutoSearch avoids
running lifecycle scripts during npm install because automatic install-time
execution is a known supply-chain risk. Plain npm install -g autosearch-ai
installs the wrapper but does not run init — use npx autosearch-ai for
the one-shot flow.
The explicit init step configures the MCP server. You should see:
✅ Python 3.12+ [OK]
✅ claude_code [Detected]
✅ MCP server [Config written]
✅ Search channels [32/39 ready]
+--------------------------------------------+
| You are all set! |
| Run: autosearch doctor — channel status |
| Run: autosearch login xhs — Chinese media |
+--------------------------------------------+
2. Open Claude Code
Restart Claude Code (or run /mcp to reload). AutoSearch tools will appear automatically.
3. Run your first search
Ask Claude:
Search arxiv for "transformer long-context 2026" and summarize the top findings.
Claude will call run_clarify → delegate_subtask → consolidate_research automatically.
4. Deep research
For longer research sessions:
Research the current state of open-source LLM inference optimization.
Use AutoSearch to search academic and developer sources, then give me a structured report.
AutoSearch handles multi-round searching, gap detection, and result deduplication.
5. Check channel status
To unlock Chinese social media channels (Xiaohongshu, Weibo, Douyin, etc.):
autosearch login xhs
autosearch configure TIKHUB_API_KEY <your-key>
See the full install guide for all options.