TaXue spent five months building Argo, a search skill that gives Claude Code, Codex and other agents 300 sources and works without an API key
Many agents search the web through one or two engines, so a question about a company returns old news and a question about a product returns advertorials. TaXue (@TaXue2025) built Argo for that: a search and deep-research tool for agents, installable as a Skill, an MCP server or a DeepSeek Harness plugin. v2.14.0 lists 300 sources and picks sources by topic and language, de-duplicates and ranks the results, and can also fetch page text, WeChat articles, PDFs and papers. Without API keys it uses only free and local sources; sources with keys join automatically. We cloned v2.14.0 (1345696) and installed only pyyaml, with no keys: 237 sources were callable on our machine; an English query returned 5 results (including the original post and blog) in about 2.8 s, and a Chinese query returned 5 Chinese results from Sogou WeChat and anysearch in about 2.7 s. We did not test deep research, logged-in search or the MCP setup.
Visit site Argo (agent search skill / MCP) ↗- Pricing
- Free
- Region
- China
- Where it's from
- Chinese community
- Added
- Nov 12, 2026
- Tested on
- Nov 12, 2026
- Version we used
- argo-search 2.14.0 (git 1345696), CLI from repo, no API keys
- Categories
- ResearchAutomation & Agents
- Original author
- 踏雪寻仙 @TaXue2025 ↗
- Our test
- Partly tested
- Cost / time
- Apache 2.0 and free. Without keys it uses only free sources; commercial sources such as Tavily, Exa or Firecrawl need your own keys and bill at their prices. The README notes a default DeepSeek budget of ¥5/month, adjustable with argo budget.
The task
Stop an agent from depending on a single search engine: switch sources automatically by language and topic so it returns fewer stale stories and advertorials.
Steps
- You need Python 3.9+. The quickest setup is to have your MCP client run npx -y github:taxueseek/argo (Claude Code, Cursor and others); for long-term local use, run the install.sh script from the README or npm install -g argo-search.
- DeepSeek Harness users can install it as a plugin: dsh plugin --profile web add argo-search, then restart dsh web.
- Try it without keys first: argo search --list-engines --routable-only shows which sources work on your machine; then argo search "your question" --max-results 5 --json, and check engines_used to see which sources were actually called.
- For better quality, add keys for the sources you want (for example Tavily or Exa) following docs/ENGINE_CATALOG.md; sources without keys are skipped.
- Fetch a single page with argo fetch "url", a WeChat article with argo wx "url", and a paper with argo paper "arXiv id".
What the author says
TaXue @TaXue2025 (X article, 10 Nov 2026): five months of solo work; v2.14.0 connects 300 sources, more than 100 of them with no configuration; ships as a Skill, an MCP server and a DSH plugin for Claude Code, Codex, Zcode, WorkBuddy, Doubao, Qwen, DeepSeek Harness and more. He argues most multi-engine search projects wrap a handful of commercial APIs and make you get the keys yourself; Argo also supports Chinese search services such as DeepSeek, Yuanbao, Zhihu and anysearch, and can search local files (skills, code, docs, notes) together with web results.
Our test
On 12 Nov 2026 on Linux (CPython 3.13.5) we git-cloned https://github.com/taxueseek/argo (1345696; pyproject version 2.14.0; Apache 2.0), created a venv with only pyyaml, set no API keys, and used the repo's bin/argo. argo search --list-engines --routable-only reported 237 implemented sources. argo search "simon willison beep locator" --max-results 5 used two free channels (parallel_free and firecrawl) and returned 5 results in about 2.8 s, including Simon Willison's blog post and his X post. argo search "Seedream 5.5 发布" --json routed to wechat_sogou and anysearch and returned 5 Chinese results (Sohu, Baidu Baike, Bilibili and others) in about 2.7 s. We did not test research, fetch, logged-in ego-search, or connect it to any MCP client.
What we found when we tried it:With no keys, Argo searched in English and Chinese using free sources and switched to Chinese sources for the Chinese query; both queries returned relevant results in under 3 s.
How to get in
The repo is public on GitHub. The original write-up is an X article in Chinese: https://x.com/TaXue2025/status/2120168395270500837 (use your browser's translate). The README has an English version (README.md).
Who it suits
People who often have agents look up Chinese sources, WeChat articles, industry news or papers, and DeepSeek Harness users who want a different search backend. If you only search occasionally and your model's built-in search is enough, you don't need it.
Pitfalls
- Without keys there are fewer usable sources (237 callable on our machine), and quality depends on which free sources respond that day.
- Fetched content is marked untrusted; an agent should still open and check sources before citing them, and a ranking score is not a measure of truth.
- Content behind logins (Zhihu, Xiaohongshu) needs ego-search with your own logged-in session; it is off by default.
- The author's claim of a large lead over other projects is his own comparison; we did not benchmark Argo against other search tools.
More in this category
- Use a web app once and get an AI write-up of how it worksResearchREAClaude CodeChromeFree tools; needs Claude or Codex◐ Partly testedOverseas
- 1,050 memory recalls audited: what makes agent memory pay offCodingNowledge MemCodexClaude CodeFree tier: 50 memories; review uses model quota○ Not tested by usChina
- Align terms, check facts, then decideCodingsinfire-skillsCodexgrill-meSkills free; needs Codex or Claude◐ Partly testedChina
- 1,169 dossiers + two evidence-based growth research methodsResearchBeRichClaude CodeIndie hackersSkill free; full paths locked in Pro◐ Partly testedChina
- Skip subagents for ordinary code search: up to +221% cost, no recall gainCodingClaude CodeTokenSaveSubagentsStudy open; running the agent still needs a plan◐ Partly testedOverseas