IMTI

Architecting, Developing, SRE, DevOps, AI/ML

MCP Is Flawed. Build With It Anyway.

Context has always been the hard problem. MCP forces you to solve it.

The Model Context Protocol has real security issues, scalability limits, and rough edges. None of that changes the fact that building custom MCP servers for internal data platforms is the right call in 2026. Here's why.

Posted by Craig Johnston Tuesday, January 20, 2026

AI Data Lake Access with MCP and S3

Building composable MCP servers for object storage

mcp-s3 exposes S3 and S3-compatible storage to AI assistants via MCP. Use it standalone or import it as a Go library to build custom MCP servers with authentication, audit logging, and content filtering.

Posted by Craig Johnston Monday, January 19, 2026

AI Data Warehouse Access with MCP and Trino

Building composable MCP servers for enterprise data

mcp-trino exposes Trino's federated SQL engine to AI assistants via MCP. Use it standalone or import it as a Go library to build custom MCP servers with authentication, audit logging, and query filtering.

Posted by Craig Johnston Saturday, January 17, 2026

kubefwd in 2026: Interactive TUI and Auto-Reconnect

The complete guide to kubefwd's new features

kubefwd now includes an interactive terminal UI, automatic reconnection that survives pod restarts and VPN drops, and a REST API for programmatic control. This guide covers all the new features.

Posted by Craig Johnston Sunday, January 4, 2026

AI-Assisted Kubernetes Development with kubefwd

Let your AI assistant manage cluster connections

kubefwd includes an MCP server that lets AI assistants like Claude Code manage Kubernetes port forwarding on your behalf. Your AI can discover cluster services, forward them as needed, and read pod logs to help debug issues.

Posted by Craig Johnston Sunday, January 4, 2026