Messy tables are everywhere: copy-pasted reports with mixed delimiters, stray whitespace, inconsistent casing, missing fields. Humans fix them in Excel; AI agents need something else — a deterministic, machine-callable tool that never hallucinates a cell value.
TableFix MCP is a self-hosted MCP server (MCP spec 2025-11-25, Streamable HTTP transport) exposing two tools:
- detect_table — analyzes raw pasted text and returns the detected delimiter, column count, confidence score, per-delimiter scores, and suspect row indexes.
- clean_messy_text — deterministically normalizes the text into clean rows (JSON), with the detected delimiter and suspect rows flagged.
On top of the server sits a simulated Alexa+ conversational web experience: type or paste messy text as if speaking to Alexa+, watch the agent call detect_table then clean_messy_text on the same live server, preview the cleaned table, and download it as a real .xlsx file — all in the browser.
Why simulated? Alexa+'s MCP add-on deployment pipeline is currently partner-only, with no public sandbox for third-party MCP servers. The hackathon rules explicitly sanction this path: a real, spec-compliant MCP server plus a simulated Alexa+ web experience. The simulation layer is clearly labeled in the UI and its source ships in the repo.
Built with: Node.js, the official MCP TypeScript SDK, Express, Vercel serverless functions, SheetJS (client-side xlsx export).
Log in or sign up for Devpost to join the conversation.