A miniapp is a small, self-contained application produced by an AI coding agent from a natural-language prompt. Unlike a full software product, a miniapp is scoped to one job: a report with filters, a calculator, a tracker, or a lightweight dashboard view. The builder describes what they need; the agent writes the application logic, interface, and any data connections.
Standalone file vs. hosted platform
Miniapps commonly ship in one of two forms:
- Standalone file: A single HTML file (often with embedded JavaScript) that runs locally or is shared as a link. Useful for quick prototypes or fixed snapshots, but usually disconnected from live data unless someone refreshes or re-exports manually.
- Hosted on a platform: The app runs inside a product environment with managed hosting, versioning, permissions, and SDK access to live APIs or warehouses. The platform handles deployment, sandboxing, and data plumbing; the builder supplies the logic.
Live data and team use
When hosted, miniapps are often wired to live data feeds: SQL against a warehouse, CRM queries, or other APIs on each load so the view reflects current state rather than a stale export. Mature implementations add versioning, roll-back, and scoped sharing so a one-off prototype can become a durable team asset.
Miniapps in GTM and analytics
Go-to-market and analytics teams use miniapps for work that does not justify a formal engineering project: channel breakdowns, campaign trackers, attribution explorers, and deal post-mortem views. The pattern replaces analyst-queue requests and one-off spreadsheets with a reusable app on shared data. See account story and forensic attribution for the kinds of questions teams often turn into miniapps.
Upside miniapps
Upside is one implementation of the miniapp pattern for GTM teams. On Upside, miniapps live in the dashboard, query healed GTM data live through the Miniapp SDK, and are built and published from Claude, Cursor, or Upside's hosted agent over MCP. Learn how Upside miniapps work →