Upsolve AI is the analytics platform for deploying governed, grounded, trustworthy data agents that know your business. Two components: Upsolve Data Agent + Agent Studio. Upsolve Data Agent acts like a best-in-class analyst: asking clarifying questions, optimizing queries, charting data, drawing insights, and creating reports. Agent Context Studio handles governance, semantic layer, context curation, with full testing, observability and evals so your agent answers the what and the why correctly.
This is the 8th launch from Upsolve AI. View more

Upsolve MCP Connections
Launched this week
Attach any MCP server to your Upsolve agent. Your agent still answers from governed warehouse data, and can now read and write where your team works: save an analysis to Notion, open a Linear ticket, pull context from docs.






Free Options
Launch Team

Memanto CloudYour agents focus. Memanto remembers.
Promoted








Hey Product Hunt, feature 3 of the Upsolve launch series.
Until now, Upsolve agents only spoke to structured sources: Postgres, ClickHouse, Snowflake, BigQuery. But once teams had agents in production, they kept asking for the next step. The answer is useful, so now put it somewhere.
Upsolve MCP Connections lets you attach any MCP server to your agent. You paste the server URL, add a one-line instruction ("use this to save analyses in Notion docs"), authenticate with OAuth, and choose which tools to allow or block.
In the demo, the agent answers "which Japanese athletes won the most medals", checks that a Sales page exists in Notion, and writes the full analysis there as a new page. The question runs as the Japan user, so row-level security still limits what data goes into the doc.
Which tool would you connect first?
choosing which mcp tools to allow or block per agent is a nice touch Ka Ling, congrats!
This is the exact seam I keep hitting with MCP — connecting is easy, deciding what the agent is allowed to do with the connection is the hard part. I like that you can allow or block tools per agent. When a write happens, like saving to Notion or opening a Linear ticket, does that go through the same governed permissions as the warehouse query, or is it up to the MCP server's own auth? Curious how you keep a read-only data agent from accidentally becoming a write-everywhere one.
the demo running as the Japan user is the detail i'd show people, row-level security usualy stops at the dashboard and then the agent pastes everything into a doc anyone can open
once the analysis lands in Notion, who can read it, only the people Upsolve would allow or anyone with access to that page??? feels like the spot where all the RLS work could leak
Plugging Notion and Linear straight into the data agent removes a lot of copy-paste for analysts. This can be a huge time saver.