Let’s talk about the engine room.
For the last 12 months, I’ve been knee-deep in AI agent builders testing, building, refining, and deploying them into real businesses.
And through all the trial and error, there are two platforms I keep coming back to:
→ n8n
→ Lindy AI
These two are the sweet spot for me balancing control, flexibility, and ease of use. They allow me to build actual agents, not just automations. Agents that take tasks, think, reason, and execute.
But what really brings them to life?
→ MCP (Model Context Protocol)
MCP is the bridge that connects agents to your real-world systems like Outlook, Slack, CRMs, you name it. It’s what allows Claude or GPT to do things, not just say things.
With MCP + Docker + Claude, I’ve been able to deploy lightweight agents that run on my local environment, process tasks in context, and deliver actual outcomes. From generating daily updates to handling meeting prep, these aren’t just fancy prompts, they’re full workflows.
And then there’s Relevance AI underrated, but powerful. Especially when you need fast custom vector pipelines, or are working with internal knowledge bases and want retrieval that’s snappy and smart.
What’s been most effective?
Here’s my personal stack for results:
Ideation + Testing: Claude Code (honestly, elite for prototyping)
Agent mind: Lindy (great for long-running agents or swarm setups)
Real-world action: MCP (connects it all)
Scaling / retrieval: Relevance AI
Most people think they need some huge platform or a 6-figure budget to get started with AI agents.
You don’t. You just need to:
→ Understand your workflow
→ Start small
→ Plug the right tools together
The best systems I’ve built didn’t come from a grand plan, they came from tinkering, testing, and improving as I went.
And now, I do this for others.
If your business is ready to explore agent workflows, reach out to me!
I help design and deploy AI systems that actually do work.
Let’s build.
- JC
JC
<100 subscribers
Support dialog
