MinusX

Agentic BI platform that turns natural-language questions into dashboards, SQL, and boardroom slides.

Freemium· Free cloud tier; paid tiers; self-hosted open-core optionAgentsClaude (BYO LLM key supported)

Best for

Pick MinusX if you're a founder or operator who wants an analyst-grade agent for your warehouse without standing up a full BI stack.

Skip if

Skip it if you already run a mature dbt + Looker/Tableau stack with dedicated analysts and strict governance requirements.

MinusX is an agentic data analytics platform that lets non-analysts ask questions of their warehouse in plain English and get back charts, dashboards, SQL, alerts, and even narrative slide decks. It currently sits at the top of UC Berkeley EPIC lab's DataAgentBench, a benchmark for realistic data-analytics tasks, and bundles ad-hoc Q&A, auto-dashboards, scheduled reports, a SQL GUI, and a knowledge base where you teach the agent your metrics and definitions.

The pitch is squarely aimed at founders, growth, finance, ops, and product teams who don't have (or don't want to bother) a dedicated data analyst. It connects to Postgres, BigQuery, Athena, DuckDB, Google Sheets, and CSV uploads, runs from a web app or directly inside Slack, and exposes an MCP server so Claude or ChatGPT can drive it as a tool. Pricing is freemium with a free cloud tier, paid plans, and a source-available self-hosted edition for teams that want their data to stay put.

Under the hood it's Claude-powered, with BYO-LLM-key supported on the cloud tier. The agentic, eval-driven approach (and the public benchmark scores) make it more credible than the average "chat with your database" tool, though like all text-to-SQL systems it still leans heavily on a well-curated semantic layer to avoid confidently-wrong answers.

Editor's take

MinusX is one of the more credible entrants in the agentic-BI race, mostly because they publish benchmark numbers and ship a real self-hosted option instead of just a chat box over your database. The Claude + MCP combo plus an actual evals/knowledge layer makes it worth a serious look for lean teams; just budget time to curate the semantic layer.

— The AI Tool Bible editorial team