The core idea: spreadsheets have been passive tools for 40 years. You put data in, formulas process it, you export it, manually move it somewhere else, repeat. We added bidirectional API integration with natural language commands, so the spreadsheet itself can send emails, update databases, orchestrate workflows across services – autonomously.
Think of it as the evolution from VisiCalc → Excel → Google Sheets (multi-user) → Sourcetable (operational). The spreadsheet stops being just a calculation layer and becomes an execution layer.
How it works:
- Natural language commands: "send emails to column A using data from column B" executes across connected email APIs
- Three integration modes: pre-built connectors (Google Ads, Shopify, Stripe, Postgres), generative connectors (AI writes the integration code in real-time for any API, Database or MCP), and a credential vault
- Full audit trail in cells – every action logs back to the spreadsheet
- Excel-compatible (formulas, A1 notation, etc.)
Technical stack: - Python/FastAPI backend with Ray for distributed execution
- React frontend with DuckDB WASM for client-side analytics
- Apache Arrow for data interchange
- LiteLLM for multi-model AI (OpenAI, Anthropic, Groq, Custom Models Based on Granite 4)
Example use case:Instead of manually exporting Shopify orders, importing to Excel, calculating reorder quantities, and emailing suppliers, the operational spreadsheet monitors Shopify, detects inventory changes, calculates reorders, and submits purchase orders to suppliers automatically. The spreadsheet becomes the workflow engine, not just documentation.
We're live at https://sourcetable.com/superagents
Blog post on the email automation example: https://blog.sourcetable.com/email-from-a-spreadsheet/
Happy to answer technical questions. We know HN will have strong opinions on whether this is the right abstraction layer for automation vs. traditional workflow tools or code.
Key decisions we made that HN will likely debate:
1. Spreadsheet as the orchestration layer (vs. Zapier/n8n/Airflow)
2. Natural language commands (vs. formula DSL or Python)
3. Generative connectors that write integration code on-demand (quality varies by API docs)
4. Excel compatibility as a constraint (limits some interactions but keeps it accessible)
Founded by Eoin McMillan and Andrew Grosser. Backed by Preston-Werner Ventures (Tom Preston-Werner from GitHub), Long Journey, Bee Partners, NextView.