D2B introduces spreadsheet infrastructure for AI agents

Summary

D2B is a spreadsheet data platform designed for AI agents, exposed through an API, MCP server, CLI, and Python and TypeScript SDKs. It ingests existing Excel workbooks in two stages: a faithful raw copy followed by typed, row-identified tables, while retaining the original values. Agents can define SQL or Python transforms instead of directly editing rows; those transforms retain lineage, recompute when inputs change, and can be exported as separate tables or written back into the original workbook without changing its styles, formulas, or other sheets. The platform stores content-addressed transforms and queryable lineage, with snapshots, named versions, operation logs, undo, and diffs showing changed transforms, rows, and downstream artifacts. Typed tables use optimistic locking through expected_version, and runtime row schemas can support constrained decoding. Column tags and role-based policies are enforced at the data layer across SQL, exports, and MCP. D2B also supports Git-based pull, review, and push workflows with three-way merges, and connects to tools including Claude Code, Codex, Cursor, VS Code, LangGraph, OpenAI Agents SDK, n8n, and Dify. Pricing is metered by write-side data operations and storage; reads are currently free, and the free tier includes a trial allocation without requiring a card. Paid plans range from $29 per month for Solo to $199 per month for Pro, with enterprise and usage-based options available.