SpacetimeDB, the reactive backend platform from Clockwork Labs, passed an integration test with flying colors but failed to convince AI agents to pick it over Convex or Supabase. A recent audit by techstackups.com used Claude Opus 5.5 to evaluate SpacetimeDB's "agent experience" (AX) across four dimensions: Awareness, Discoverability, Reputation, and Integration. While the model successfully built a live-updating app without human help, it consistently recommended competitors for web applications, scoring a mediocre 3/4 overall.

The Awareness Gap

The audit revealed a significant disconnect between SpacetimeDB's 2026 rebrand and the model's training data. When asked about "Spacetime" or its new tagline, "Let AI ship the whole backend, glue-free," Claude Opus 5.5 admitted it didn't recognize the product, confusing it with Convex. The model's knowledge was frozen in 2025, correctly identifying SpacetimeDB as a game backend but remaining unaware of the push into general web apps, the C++ support, or the newly released agent tooling like the official MCP server and llms.txt.

Discoverability and Reputation Struggles

In discoverability tests, SpacetimeDB appeared as a top pick for multiplayer games but consistently ranked second to Convex for web apps. The audit found that agents struggled to find SpacetimeDB's official agent tools via search, often landing on community MCP servers instead. Reputationally, the agent was skeptical of SpacetimeDB for standard SaaS dashboards, citing limited SQL capabilities and a lack of public production evidence beyond the game BitCraft. One agent even cited SpacetimeDB's own blog post, "Why You Should Choose Convex," as evidence against using SpacetimeDB for collaborative web apps.

Integration Success

Despite the marketing and discoverability issues, the technical integration was flawless. In a fresh Claude Code environment, the agent installed SpacetimeDB, built a working live page, and correctly handled private data without any human intervention or additional plugins. This demonstrates that while the AI agents don't know SpacetimeDB exists as a web platform, they are highly capable of building with it once pointed to it.

Key Takeaways

  • Training Data Lag: AI models are unaware of SpacetimeDB's 2026 rebrand and web app focus, relying on outdated 2025 game-backend knowledge.
  • Competitor Dominance: Convex and Supabase are the default recommendations for web apps, with agents citing Convex's superior developer experience.
  • Search Visibility Issues: Official agent tools like the MCP server and llms.txt were difficult for agents to find via web search, often overshadowed by community alternatives.
  • Technical Capability: Agents can build complex, live-updating apps on SpacetimeDB with zero human assistance, proving the platform's robustness despite poor discoverability.

The Bottom Line

SpacetimeDB is technically superior for reactive backends but invisibly bad at marketing to the very agents it claims to serve. Until Clockwork Labs fixes its SEO and ensures its agent tooling is indexed correctly, AI developers will keep choosing Convex for web apps despite SpacetimeDB's superior integration capabilities.