By name: lexai-research description: Use the LEX AI tools to research regulatory and compliance topics. Chain keyword search with related-post lookup to surface adjacent material a keyword search alone would miss.
Researching with LEX AI
You have three LEX AI tools (provided by the lexai MCP server). Use them in this order for any regulatory/compliance research question.
1. search_fts(query, ...): start here
Keyword search across LEX AI content (legal updates, summaries, regulations, acts). Use the user's terms as-is, or quoted phrases for multi-word concepts (e.g. "data protection"). A bare term also matches posts carrying it as a hashtag (e.g. aml surfaces #aml-tagged posts). Returns up to 50 posts with title, url, type, publish_date.
Optional filters:
org: organization name (e.g."BaFin") — returns only posts from matching organizations.region: country/region name or ISO code — e.g."Germany"/"DE","United States"/"US","European Union"/"EU". United Kingdom is"United Kingdom"/"GB"/"UK"; China is"Mainland China". Unknown values return nothing.post_type,start_date,end_date,page_size(≤50),page_number.
If results are sparse or feel narrow, immediately run step 2; don't ask the user.
2. graph_neighbors(urls, top_k?, min_similarity?): expand
Pass the top 3–5 url values from step 1. Returns related posts the keyword search missed. Default top_k=10, min_similarity=0.7. Lower min_similarity to ~0.55 for exploratory research.
3. get_post_by_url(url): fetch full content
When the user asks "summarize that post" or "what does
Output
When citing a LEX post in your reply, always include the url from the tool result so the user can open it in LEX. Don't summarize lists of titles without urls; the citation is the value.
Don'ts
- Don't call
graph_neighborswith an emptyurlsarray; runsearch_ftsfirst. - Don't ask the user to refine their query before chaining search to related-posts; that's free signal you can gather yourself.
- Don't paginate beyond
page_number=2forsearch_ftsunless the user explicitly asks. Keep responses tight.