Extending H2O with APIs, SDKs, and MCP Integrations
- Agentic AI
- Enterprise h2oGPTe
- H2O Driverless AI
This lesson walks through the programmatic extensibility available across the H2O platform for developers who need more control than the UI provides. It covers Python SDKs and REST APIs for every component, hosted Jupyter Labs with sample notebooks, and how this scriptability supports CI/CD pipelines for automated model deployment. It covers h2oGPTe's APIs for collections and document ingestion, documented through Swagger UIs that generate client code in several languages. It closes with Driverless AI custom recipes and the Model Context Protocol, which lets agents call external tools such as Salesforce or GitHub through MCP servers.
Key takeaways
- Why every H2O component being scriptable through APIs supports CI/CD pipelines for automated deployment
- How h2oGPTe's Swagger-documented APIs let you generate client code in Python, JavaScript, or Go
- How Driverless AI custom recipes let you inject your own feature transformations or scoring functions
- What the Model Context Protocol enables: agents calling external tools like Salesforce or GitHub through MCP servers
Start owning your AI