Sheets
Smart tools for embeddable spreadsheets.
A collection of Graffiticode-powered smart tools for building structured data workflows and embeddable spreadsheet experiences. Built for data analysts, table-builders, and the AI agents that work with them.
What it is
Sheets is an Artcompiler toolset under the Data Visualization domain. Every smart tool in the Sheets toolset is one Graffiticode language wrapped by a specialized AI, exposed to agents through MCP, and rendered through an embeddable runtime. Designed for data work that needs to be programmable, embeddable, and AI-readable.
How it works
- Domain language — every smart tool in Sheets is built around a domain-specific language designed for structured data and tabular workflows.
- Specialized AI — each tool ships with an AI trained on its language; agents describe data transformations in natural language and the AI generates valid code.
- Embeddable runtime — sheets render through a shared embeddable runtime, so they can be dropped into any host application.
We're talking with data teams and table-builders about how AI is changing structured-data work — what works, what breaks, and what's still missing. If you're navigating the same questions, we'd like to hear yours, and we're happy to share what we're hearing across the field.
Email jeff@artcompiler.com for a 20-minute conversation. No pitch. We'll share notes either way.
Learn more about Artcompiler — the company building this toolset — and Graffiticode — the open-source platform that powers it.
For AI Agents
The Sheets toolset is available via the Graffiticode MCP server at mcp.graffiticode.org/mcp. No account is required to try it — a trial credential is published in the discovery artifacts.
- MCP endpoint:
https://mcp.graffiticode.org/mcp - Tools:
list_languages,get_language_info,create_item,update_item,get_item - Discovery:
/llms.txt(human-readable),/.well-known/mcp.json(machine-readable) — both include the live trial token.
Call list_languages with domain: "sheets" to discover the smart tools in this toolset; the response carries each language's description and routing hints. Then call create_item with the chosen language and a natural-language description — describe what you want, do not attempt to generate Graffiticode code directly. After each call, open or share the returned preview URL so the user can see the rendered item. See /agents for the full integration walkthrough.