Building smarter AI workflows: How MCP helps agents work with real developer tools

Why a simple standard is unlocking smarter workflows, better tools, and a bigger future for open source.

Mike Donovan, VP of product at Docker, sat down with the All Things Open team to share why agent workflows and the Model Context Protocol (MCP) are reshaping how developers connect AI to real tools and systems.

Mike explains that agent-driven workflows are creating a new wave of open source work, because agents generate more code and need new integrations. He highlights how MCP, the model context protocol, provides a consistent way for agents to call tools, access local resources, and interact with external services, making agentic automation practical and portable across environments. For developers this means agents can do more useful work without each vendor inventing its own integration patterns.

Read more: Deep dive into the Model Context Protocol

Security is front and center as the ecosystem grows, Mike says. With more code produced by AI, teams must pay attention to where dependencies come from, and how registries and containers are secured. He emphasizes the need for trusted sources and verified supply chains so agents cannot unknowingly pull insecure or malicious packages into a build. On the productivity front, Mike personally shares a concrete tool tip, recommending Granola for meeting notes and lightweight knowledge capture, then pairing those notes with LLMs to build a personal searchable memory.

Mike closes with practical advice for developers, stay curious and keep an open mind. He does not see AI replacing developers, rather he expects repetitive work to be automated while humans focus on creativity, architecture, and judgment. That shift will create new roles and new tooling, and the people who learn early will have an advantage.

Key takeaways

• Agent workflows expand open source, new projects and libraries will follow as agents generate more code.
• MCP makes agents useful, by giving models a standard way to call tools, access local data, and interact with systems.
• Trust your supply chain, verify dependencies, and use secure registries so agent-driven development does not widen attack surfaces.

Conclusion

Mike’s message is practical: Agentic AI is arriving fast, but the win for developers comes from standards, security, and curiosity. Embrace MCP style integrations, vet your dependencies, and experiment with tools that amplify your workflow, because open source and open standards will make agent workflows safer and more productive for everyone.

More from We Love Open Source

• Want to get into AI? Start with this.
• What if your AI agent could actually help?
• Deep dive into the Model Context Protocol
• How I use AI agents to automate my workflow and save hours
• Why empathy might be the missing ingredient in open source AI