Keenable SELECT: an agent that searches the web in SQL
Source Entity
Hacker News
Keenable SELECT is a novel AI agent that utilizes SQL-based queries to perform efficient web searches. By streamlining data retrieval through DuckDB, it minimizes token usage and enhances transparency in research reporting.
The Evolution of Web Search: Understanding Keenable SELECT
The landscape of AI-driven information retrieval is undergoing a significant shift with the introduction of Keenable SELECT. Unlike traditional search engines that provide agents with a list of links—forcing the model to consume expensive tokens to read and synthesize disparate web pages—Keenable SELECT functions as an MCP (Model Context Protocol) server that treats the web as a database. By utilizing a singular, read-only DuckDB SELECT statement, the system directly queries live web data, fundamentally changing how agents interact with the internet.
Mechanism: The SQL-Centric Approach
At its core, Keenable SELECT operates on the premise that structured data is more efficient to process than unstructured text. The system executes web and semantic operators outside of DuckDB, injecting the resulting data back into a row set that the SQL engine then processes. This architecture allows the agent to perform complex queries, such as tracking the movement of AI researchers between frontier labs since 2025, with surgical precision. By bypassing the need to 'read' numerous full-page documents, the system reduces the latency and computational overhead typically associated with LLM-based research agents.
Transparency and Auditability in AI Research
One of the most compelling aspects of Keenable SELECT is its commitment to transparency. Every report generated by the tool comes with a 'trajectory'—a detailed log of every query, tool invocation, and raw result set. This allows users to verify the provenance of information, a critical requirement in an era where AI-generated content is often criticized for its lack of explainability. By linking the finished report back to the specific SQL operations, Keenable SELECT provides a verifiable chain of custody for its data.
Broader Implications for Data Intelligence
The shift toward SQL-based web searching marks a move toward 'Data Intelligence' rather than mere 'Information Retrieval.' As AI agents become more autonomous, their ability to interface directly with structured outputs will define their utility in professional research. Keenable SELECT demonstrates that by standardizing the interface between the web and the database, we can achieve more reliable, faster, and more cost-effective research outcomes.
Future Trends in Agentic Workflows
Looking forward, the success of Keenable SELECT suggests that we will see a decline in 'black-box' search agents. Future iterations of autonomous agents will likely favor architectures that emphasize tool-use efficiency and SQL compatibility. As organizations demand higher accuracy from their AI tools, the ability to perform complex, multi-step queries without ballooning token costs will become the benchmark for industry-standard AI research assistants.