Moving a spreadsheet into a database usually means losing the people who lived in it. Here's the version where your team keeps their table, their formulas, and their views — and the file rituals stop.
With a coding agent (Claude Code, Codex, Cursor) driving tabledi, the migration is a conversation, not a project:
Columns are typed on import. Your formulas survive — VLOOKUP, XLOOKUP, SUMIF and 60+ others are built in with spreadsheet semantics. The engine is columnar and holds millions of rows; aggregates that froze Excel come back in milliseconds.
Because Postgres solves the size problem and creates a people problem: the ops teammate who owned that spreadsheet can't open psql. tabledi keeps both sides of the handoff working — your agent gets a CLI, an MCP server, and deterministic, audited writes; your teammates get invited into the workspace and see the same tables and views, updating live, formulas intact. No exports, no stale screenshots — and no one has to learn SQL to check a number.
Every write is validated against the column types, optimistic-locked, rate-limited and audited. Every query answer carries a generation stamp — ask the same question a hundred times and you get the same number a hundred times. And the whole thing is a single Rust binary you can self-host, so "move it out of Excel" never has to mean "move it out of the building".
Install the CLI, hand it to your agent, point it at the file.
Get started See what it does