SpacetimeDB AX Audit Finds Strong Agent Integration but Weak Web-App Positioning

Summary

Tech Stackups audited SpacetimeDB through the experience of an AI coding agent, testing awareness, discoverability, reputation and integration. The audit used Claude Opus 5.5 in Claude Code 2.1.285 on October 7, 2026; each prompt ran once in fresh environments, so the results represent one developer's experience rather than a benchmark. The model understood SpacetimeDB as a database-and-server platform for multiplayer games, but did not know its 2026 rebrand, web-app focus or official agent tooling from memory. With web search, it found SpacetimeDB quickly for multiplayer games, while Convex remained the preferred option for collaborative web apps and agent-built applications. Once the product was named, the model generally described its capabilities accurately, but it recommended Convex or Supabase for typical web apps, citing concerns including limited SQL functionality, webhooks, authentication, ecosystem size and production evidence. Some claims were wrong because the agent relied on an unofficial repository mirror, including the latest version and file-storage capabilities. The integration test was substantially stronger. Claude Code installed SpacetimeDB 2.10.2 locally in 4 minutes 28 seconds without signup or an API key, loaded 50 customers, 30 products and 300 orders, and answered a revenue question correctly. Because SpacetimeDB SQL could filter, join and count but could not add, group or sort, the agent copied current data into SQLite for reporting. It then built a live revenue page, changed an order to refunded, and observed the page update in about 23 milliseconds without a reload. It also made the customer table private and exposed a public view containing only IDs, names and regions, preventing anonymous clients from receiving email addresses. The agent did not use SpacetimeDB's MCP server, plugin, skills or llms.txt; it relied on generated project rules, package types and source code. The audit gives awareness, discoverability and integration 3/4, 3/4 and 4/4, but reputation 2/4, for an overall 3/4. It concludes that SpacetimeDB is usable by agents, while its newer web-app positioning and agent tooling remain difficult for them to discover.