

Redefining Developer Workflows in the AI Era with MCP | Steve Manuel
6 snips Oct 7, 2025
Steve Manuel, Founder and CEO of Delibso, discusses the groundbreaking MCP protocol that ensures safe AI integrations. He explains how mcp.run allows developers to utilize shared servers without losing control, making experimentation safer. Manuel highlights the importance of plugin-safe AI for developer tools, the risks of untrusted MCP servers, and emphasizes that first-party code remains crucial in the AI landscape. Additional insights include automating workflows and the potential of multimedia interactions, redefining how applications function in the AI era.
AI Snips
Chapters
Transcript
Episode notes
Dylibso's MCP Origin Story
- Dylibso shifted from WebAssembly plugin work into MCP after seeing the protocol's potential.
- They quickly implemented MCP on top of their isolation stack and built mcp.run and Turbo MCP.
AI-Native Plugin System
- MCP aligns with the trend that all software will become AI-enabled, making a common plugin system crucial.
- Treating MCP as the plugin system for AI unlocks broad extensibility across applications.
MCP As A Universal Tool Interface
- MCP standardizes tool interfaces so any client can use the same server implementation.
- This enables write-once, run-anywhere tool reuse across models and agents.