FR EN

07 November 2025

BDSJ - Brained DeepSearch jan

BDSJ is an agent that alternates between searching (internet and/or RAG) and reasoning to respond to complex queries, “x4” means that it has 4 cycles.

Easy customization: It’s easy to modify it to suit a research field, your use, an LLM, or other purposes.

Among the MCPs of interest for research are: Paper Search, Searxng simple MCP (without scraper), SearXNG MCP (with scraper), Jina (paid), Tavily (paid), Brave (paid).

Jina, Brave and Tavily are paid services, but they allow you to avoid overloading the context with scraped (fetched) web pages.

It is often very useful to add a RAG search tool because vector-based search is much more effective than web search.

Prompt System

Its prompt system is in English because that’s generally the language that LLMs understand best. if you speak another language besides English, for best results, I recommend writing your queries in English and translating the response, or you can modify an instruction at the end of the system prompt so that it writes the response in your language:

Write a structured response that:

Usage Tips

With some LLMs, it’s best to start with: “Follow your system prompt’s process to write a … (guide, article, report, study, analysis, detailed comparison, etc.)”

You can give it a URL and ask it to scrape (fetch) the page (scrape this URL: the_url.com). This function (fetch or scrape) is available with many MCPs.

Sharing

If you modify it and share it elsewhere, please share the link here or in the domains pages if it’s specific to a domain.