streamboard

streamboard turns natural-language requests into hosted dashboards. Your AI assistant authors a KPI / chart / table spec through an MCP tool; your code pushes live values into bindable slots from any runtime — Node, Bun, Workers, or shell.

How it works

1. Author — ask any MCP-aware LLM (Claude, Cursor, Codex, ChatGPT) to “make me a streamboard with three KPIs and a weekly-revenue area chart”. The MCP tool at mcp.usestreamboard.com writes a versioned json-render spec.
2. Render — the resulting dashboard lives at https://usestreamboard.com/s/<id>. Public or org-scoped. Palette-customisable. Append-only versioning.
3. Push — mint a per-board data token in the web app, wire your worker (cron, agent, telemetry pipe) to @streamboard/sdk, and stream fresh values into the bindable slots. No LLM cost per refresh.

CLI + MCP

streamboard ships a CLI for direct read / push / pull / codegen access, plus an MCP server that every major AI tool can connect to.

npx streamboard streamboards ls
npx streamboard streamboards push <id> --state '{"kpis":{"mrr":{"value":"$48k"}}}'

For clients that support MCP natively (Claude Desktop, Cursor, Codex, VS Code, JetBrains, Windsurf, Antigravity), connect the MCP server at https://mcp.usestreamboard.com/mcp.

Next steps

• Quick Start — sign in, connect an MCP client, author your first streamboard
• Live data + tokens — mint a token + push values from a worker
• Bindable slots — author specs with { $bind: "field.path" } refs
• Versioning — append-only history, pinning, restoring
• Integrations — Claude Code, Claude Desktop, Cursor, Codex, ChatGPT, VS Code, JetBrains, Windsurf, Antigravity