$ loading_
Run traceable read-only queries across sources with cited answers or typed refusals.
Copy the install command and let the AI configure it · recommended for beginners
No copy-paste install info for "RecoSearch" yet — see the docs or source repo.
Query the top 10 refunded orders in the last 30 days across the order and refund systems. Return order ID, refund amount, and refund reason. Cite every row with source table name, primary key, and raw fields; if the data is insufficient, return an explicit refusal.
A query result with row-level provenance, or a structured refusal if evidence is insufficient.
Using the customer feedback database, support tickets, and product defect table, identify the top three issue themes from the last quarter and provide cited evidence rows for each theme. Do not infer anything unsupported by source records.
A summary of issue themes with evidence citations, ensuring every conclusion is traceable to source data.
Answer whether monthly active users are higher than last month, using only the analytics fact table and daily active summary table. Return the conclusion, calculation basis, and citations for every data row; refuse if the metrics are inconsistent.
A contract-enforced answer with full provenance, preventing unsupported model hallucinations.
Aggregate MCP servers and route tools intelligently for efficient parallel work.
Search the web and read single or multiple pages efficiently.
Search and cross-reference prior art across papers, patents, books, and standards.
Validate MySQL and Snowflake data integrity during migrations and ETL workflows.
Break down complex research questions and produce structured, credibility-scored reports.
Turn unstructured documents into a searchable knowledge base for AI agents.