spool

Inspiration

What it doesSPOOL

SPOOL is a local-first, agent-assisted data migration system designed to turn messy CSV files into validated, structured, database-ready outputs without forcing the user to manually supervise every transformation.

The user provides a source file and selects the desired outcome. SPOOL then runs a deterministic workflow that profiles the data, infers field types and mappings, creates a transformation plan, dry-runs it, identifies unsafe ambiguities, executes the migration, verifies the result, and produces an export.

Its key WebMCP innovation is Temporal WebMCP. Instead of exposing every tool to an AI agent permanently, SPOOL dynamically exposes only the capabilities that are valid for the current workflow state. For example, planning tools are available only after profiling is complete, execution tools appear only when the migration is ready, and stale capabilities disappear when mappings or source state change.

This means the website—not the model—owns workflow correctness.

SPOOL uses constrained deterministic transformations rather than arbitrary AI-generated code. It includes typed schema inference, data validation, revision-aware mappings, dry runs, Worker-based execution, recovery after refresh, violation reporting, and spreadsheet formula-injection protection.

The product is designed around a simple experience:

Add data → choose outcome → run Autopilot → resolve only genuine ambiguities → receive verified output.

The broader idea behind SPOOL is that complex web workflows should expose agent capabilities contextually rather than giving agents a permanent toolbox and expecting them to manage application state correctly themselves.

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for spool

Built With

• deterministic-transformation-pipelines
• indexeddb
• javascript
• local-first-storage
• schema-inference
• vercel
• web-workers
• webmcp-dynamic-tool-registration