Point at your data.
It's alive in chat.

Drop a messy CSV on abagraph, type one sentence into Claude, and a real, interactive app streams itself into existence inside the chat window — a table you can sort, a chart that redraws, filters that respond. Keep talking and it reshapes live. There was never a schema. There was never a frontend. There was never a deploy.

feel it in 10 seconds bring your own CSV

the joyless work just disappeared

No database to provision. No migrations to plan. No ORM to wire. No dashboard sprint. Your data becomes provenance-rich facts the moment it lands, and the interface generates itself in the chat you already use.

trust is a click, not a promise

"Chat with your data" tools fail silently — a plausible number that means the wrong thing. Here every cell carries its provenance: click any value and see the exact source row it came from, its confidence, and its full history. Facts land provisional and an async judge raises confidence in front of you — watching the chips flip from provisional to confirmed is the product.

alive, not rendered once

• Streams in — rows and bars build themselves on screen, not a static card.
• Reshapes on every turn — "only Q2, by team" re-lays the same app out live.
• Keeps living — confidence updates arrive without you asking.
• Remembers — it's a bitemporal store underneath: ask what changed, time-travel to any moment.