Anthropic gives Claude connector devs their own metrics dashboard

If you have built tools for Claude, you probably know the frustration of shipping an MCP connector into a complete black box with zero visibility into who is using it or when it breaks.

Anthropic just rolled out two practical features for connector developers: an observability dashboard and in-app directory submissions.

Inside Claude's new developer dashboard

The new dashboard lets developers monitor how their published connectors perform across Anthropic's lineup. You can track active users, total tool calls, and directory rank over time, alongside core health metrics like error rates and latency.

It also pinpoints failures at a per-tool level and breaks down usage across different environments, so you can see whether people are triggering your tools in Claude, Claude Code, or Cowork.

The dashboard is currently in public beta for Team and Enterprise admins. Along with this update, developers can now submit their MCP servers directly within Claude instead of jumping through external hoops. Anthropic notes that the directory now hosts over 300 third-party connectors used by millions of people daily.

Why it matters

Anthropic is turning MCP from an open experimental protocol into a professional developer platform.

For builders, having real usage metrics and latency breakdowns makes maintaining an AI agent tool feel like operating actual software infrastructure rather than a side project. You can finally see what fails and double down on what works.

It also sends a clear signal to rivals like OpenAI: the next battleground is developer retention. By giving tool builders real distribution and proper observability, Anthropic makes it much harder for developers to walk away.