One MCP endpoint gives the agent you already use data connectors, live dashboards, shared memory, and an analysis loop built for depth.
Add MCP to your agent
I want to start using phaide.ai. Here are the steps: First, sign up at https://beta.phaide.ai/signup and complete onboarding and organization creation. Then add the MCP server (https://api.beta.phaide.ai/v1/mcp/) and complete the OAuth authentication. Use the phaide_get_context tool to start the setup.
Quick walkthrough
Low-cognitive-load HTML reports, generated with a share link. Constrained plot libraries and styles keep quality consistent across every harness and LLM.
ChatGPT vs Claude Code + Phaide
A controlled comparison — every report was generated from the same dataset with the same prompt: Which feature should we improve first to drive revenue growth?
See the dataset ↗
Over 750 data connectors sync straight from the SaaS tools and databases you already run. No pipeline? You don’t need to build one.
Browse supported data sourcesIt finds the incidents you haven’t noticed and the quiet customer complaints — overnight. Scrolling the discovery feed on a break should be fun. (Unless the news is bad.)
Same wall-clock time, deeper findings. Your agent can call a fast hypothesis-testing loop purpose-built for data analysis.
Phaide learns your semantic layer and business context from conversations, and shares it across the organization — a mistake one agent makes is never repeated by the team.
It keeps agents from going rogue. Every query against every data source is logged and monitored, and a supervising model blocks dangerous queries using the policies you set.
Beautiful dashboards your team can rearrange by drag and drop. Unlike a static HTML export, every value and chart updates in real time.
“We believe a great data visual carries a message. To keep generated visuals both consistent and flexible, we redesigned the plotting tool itself.”
For individuals getting started.
For growing teams.
For organizations at scale.
For advanced security & scale.
A full setup guide for connecting Phaide to Claude — or any MCP client. Paste one URL, approve the connection in your browser, and start asking questions of your own tables from the chat you already use.
Creating a read-only database user is not enough when you let an AI read production data. Using PostgreSQL as the example, here's how to put a Query Broker between the AI and the database — and narrow both the data it sees and the work it can run.
From open-source quelmap and Lightning-4b to Phaide AI — autonomous data exploration, automatic masking, dashboards, and documents. Launching today on Product Hunt.
Learn what Agentic BI is, how it differs from Tableau, Power BI, and Metabase, and where Hex, Omni, and Basedash fit in.
Traditional BI and AI copilots only work when someone already suspects a problem. See why agentic analytics shifts data teams from reactive querying to autonomous discovery.
A structured guide to AI-powered data analysis — from the fundamentals to real business use cases, the benefits and challenges of adoption, and the leading services.
Give Phaide a name and a goal in plain language, and an AI agent writes, tests, and assembles a live, shareable dashboard from your own data.
Phaide masks sensitive columns the moment data is imported, and catches sensitive values the instant you type them in chat — so the AI sees patterns, never the people behind them.
Claude, ChatGPT and Perplexity are products of their respective owners.
Connect the MCP endpoint and ask your agent a real question about your data.
