AutoSearch plugin
AutoSearch gives an agent search over 40 channels, from arXiv to Chinese social media, and returns sources with links.
- 55 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The engine queries external services with your keys and for some channels stores account logins on your machine.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/autosearchDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The engine queries external services with your keys and for some channels stores account logins on your machine.
This is third-party code. Review the repository files before installing.
What it does
AutoSearch is open-source deep search infrastructure for agents. The engine queries 40 channels: academic databases such as arXiv and PubMed, GitHub, DuckDuckGo, Hacker News, YouTube and about ten Chinese sources including Bilibili, Zhihu, Weibo and Xiaohongshu. Results are merged and deduplicated, and each carries a source link. The engine makes no language model calls: your own agent assembles the final answer with its model and prompts. It connects as a Claude Code plugin, an `autosearch-mcp` MCP server or through `npx autosearch-ai`. `autosearch doctor` shows which channels are ready and which need a key or login.
Who it is for. Researchers and developers who use an agent for surveys across academic databases, forums and Chinese platforms without collecting links by hand.
Good fit when
- Fresh arXiv papers on a topic are needed along with the discussion around them on Hacker News.
- Similar open-source projects on GitHub must be found that general search misses.
- Opinions on English forums and Chinese social media are being compared.
- The agent needs a source URL for every fact rather than a recap from memory.
Not a fit when
- A finished report is wanted: the engine returns material and your agent draws the conclusions.
- One web search is enough and 40 channels only complicate setup.
- Xiaohongshu and Xueqiu data is needed without account login or a service key: some channels open only after configuration.
Example request
Use AutoSearch to find fresh arXiv papers on agent memory and compare them with the Hacker News discussion, links required.Limitations
Some channels work without setup, others need keys such as `YOUTUBE_API_KEY` or `TIKHUB_API_KEY`, and Xiaohongshu and Xueqiu need a login through `autosearch login`. `npx autosearch-ai` runs initialisation only when you invoke it. An `install.sh` script is also provided piped through `curl`, which is worth reading before running. The engine calls no model, so over MCP your agent needs its own model provider key.
How to disable. Remove the plugin with `/plugin uninstall` or the `autosearch` entry from MCP settings. Erase saved logins and keys separately and remove the `autosearch-ai` package through npm.
FAQ
How many channels work without keys?
Academic databases, GitHub, DuckDuckGo and Hacker News work without setup. YouTube, Bilibili and Twitter need keys, and `autosearch doctor` lists them.
Which agents connect?
Claude Code, Cursor, Cline, Zed and other MCP clients. There is also a dedicated Claude Code plugin.
Does the engine write the answer?
No. It returns ranked results with links and your agent builds the final answer with its own model.
How do I enable Xiaohongshu?
Run `autosearch login xhs` or set `TIKHUB_API_KEY`. Then check the status with `autosearch doctor`.
Does anything run on `npm install`?
No, install scripts do not run. Initialisation happens only after an explicit `npx autosearch-ai` or `autosearch init`.
Discussion
No questions or reviews yet.
Related
PluginsLocal web search for an agent with no API keys: results are ranked on your machine and only the relevant passages come back
A skill for Claude Code and Codex: one query goes to 52 search channels and the results merge into one deduplicated JSON
A Doubao-based web search MCP server for agents with Chinese models: detailed excerpts and publish times to the second.
Self-hosted web search gateway for agents: MCP server, HTTP API and library with fallback across Tavily, Brave, Exa, SearXNG
A local MCP server with no API keys: multi-engine web search, page fetching and document reading tuned for LLMs.
A deep research MCP server: multi-source search, result fusion and cited synthesis through a Qwen3 model on OpenRouter
README in short
A channel table with setup for each, three install paths, `autosearch doctor` output and links to MCP and channel docs.