Capstead

I built Capstead: A Governance & Observability Layer for AI Capabilities in Spring Boot

As more teams integrate AI into their Spring Boot applications, I kept running into the same questions:

• Which AI capabilities does our application expose?
• Who owns each capability?
• How much does each capability cost?
• Which capabilities are failing or exceeding latency budgets?

Frameworks like Spring AI provide excellent model integrations and metrics, but I wanted visibility at the business capability level rather than just individual model calls.

That's why I built Capstead, an open-source governance and observability control plane for AI capabilities in Spring Boot.

With a simple @Capability annotation (or a bodyless @CapabilityClient interface), Capstead provides:

• 📋 Live capability catalog
• 💰 Per-capability cost attribution, token usage, latency, and success rate
• 🎯 Daily budget enforcement
• 🌳 Durable execution history with parent/child execution trees
• 🤖 MCP export so capabilities can be exposed as agent tools
• 🔌 Provider-neutral support (Spring AI, LangChain4j, custom SDKs, or any HTTP client)

The goal is to provide a governance layer on top of existing AI integrations without changing how developers build AI applications.

I'd love feedback from engineers building AI-enabled systems:

• Is this a problem you've encountered?
• What governance or observability features are you missing today?
• What integrations would you like to see next?

GitHub: https://github.com/satya-anguluri/capstead

I'd really appreciate your thoughts and suggestions!