Shelby Web OfficialAugust 2026

OpenCode Changed the Way I Build Websites:
Why AI Coding Tools Are Becoming Essential for Small Web Agencies

By Dave Richardson — Founder / Lead Developer, Shelby Web Company

There's something interesting happening in web development right now. AI isn't simply helping developers write code faster. It's changing the way developers think about building software altogether.

Over the past year, I've experimented with different AI-powered coding tools, development environments, models, and workflows. Some were incredibly impressive. Some were frustrating. Some made me wonder how I ever built websites without them. But eventually, I ran into a problem that had nothing to do with the quality of the AI. I was spending too much time managing the tools instead of building. Then I discovered OpenCode. And it changed my development workflow in a way I honestly wasn't expecting.

AI-assisted web development workflow

First, What Is OpenCode?

OpenCode is an open-source AI coding agent designed to work directly with your development environment. It's available through a terminal interface, desktop application, and IDE extension. More importantly for developers who work with multiple AI providers, OpenCode isn't locked into a single model provider. Its documentation describes support for a wide range of providers and models, including the ability to use local models. That flexibility is one of the things that immediately caught my attention.

Instead of thinking: "Which AI application am I going to use?" I could start thinking: "Which model or provider makes the most sense for this particular job?" That's a fundamentally different approach.

The Problem With AI Coding Tools Nobody Talks About Enough

When AI coding tools first started becoming a serious part of my workflow, it was easy to get excited. You could ask an AI to make a change to a website and watch it happen. You could ask it to troubleshoot an error. You could have it inspect files. You could describe a feature in plain English and have it help implement it. And the free tiers made everything feel even better. At first, there seemed to be plenty of room.

Then you hit the limits. Once you started paying for subscriptions, another problem could appear: the amount of usage available wasn't always enough for the kind of development work I was actually doing. I found myself running out of credits during development sessions. Sometimes, one substantial round of edits could consume a surprising amount of usage. So I started doing what seemed like the logical thing. I switched tools. Then switched models. Then switched modes. Then switched providers. Then moved back to the original environment. The work itself wasn't suffering. The workflow was.

I Wasn't Cutting Corners. The Corners Were Cutting Me.

This is an important distinction. I wasn't using AI to lower the quality of my work. I wasn't telling an AI: "Just get it close enough." If something needed to be fixed, I wanted it fixed correctly. If a client needed a feature, I wanted the feature implemented correctly. If a website needed an improvement six months after launch, I still wanted to make that improvement.

The problem was that getting the work done could take exponentially longer because I was constantly working around limitations. I'd have to re-establish context. Move between environments. Explain the project again. Try another model. Deal with another usage limit. Then pick up where I left off. The corners weren't being cut. The corners were cutting me. And that taught me something important about AI-assisted development.

Faster Doesn't Mean Worse

There's still a misconception that if AI allows a developer to accomplish something much faster, the developer must somehow be cutting corners. I don't agree with that. If a task used to take four hours and a better tool allows me to complete it properly in one hour, that doesn't mean I should intentionally make it take four hours just so it looks like I worked harder. The question shouldn't be: "How long did this take?" The question should be: "Was it done correctly?" That's the standard that matters.

If a project needs something, it needs something. It doesn't matter whether that need appears during the initial development phase or six months after the website launches. A business doesn't stop evolving because a development project has technically been marked "complete." That's especially true with websites. Businesses change. Services change. Customers change. Marketing strategies change. Google changes. Technology changes. And sometimes a business owner simply looks at their website one day and says: "You know what would make this better?" If it makes sense for the business, I believe it deserves to be considered.

That's Where OpenCode Became Interesting

OpenCode gave me a different way to approach AI-assisted development. Instead of treating AI as a separate website where I went to ask coding questions, I could work with an AI coding agent much closer to the actual project. OpenCode can analyze a project, understand its structure, work with files, and assist with development tasks directly. Its initialization process can create an AGENTS.md file containing project-specific information to help the agent understand the codebase and its conventions.

That might sound like a small feature. For a developer, it isn't. Context matters. A lot. The more an AI system understands about the project it's working on, the more useful its assistance can become.

The Ability to Choose Your AI Model Matters

Another major reason OpenCode stood out to me is its provider flexibility. OpenCode's current documentation says it supports 75+ LLM providers through its provider ecosystem, and it can also work with local models. That doesn't mean every developer needs 75 AI providers. It means you're not building your entire development workflow around the assumption that one company, one model, or one subscription will always be the best option. That's important in an industry changing as quickly as AI.

Today's best model may not be tomorrow's best model. Pricing changes. Models improve. New models appear. Providers change their policies. Usage limits change. Having flexibility means your development workflow can evolve along with the technology.

AI Doesn't Replace the Developer

This is probably the most important part of this entire discussion. OpenCode doesn't magically turn somebody into a professional developer. AI can write code. That doesn't mean the code is automatically correct. AI can make changes. That doesn't mean those changes are automatically appropriate. AI can find a solution. That doesn't mean it's necessarily the best solution. A developer still needs to understand what they're building. They need to understand the client's business. They need to understand the existing codebase. They need to test. They need to review. They need to recognize when something doesn't make sense. And they need to take responsibility for the final result. That's why I don't look at AI as replacing what I do. I look at it as multiplying what I can do.

Think of It as a Development Partner

The easiest way I can describe AI-assisted development is this: Imagine having another developer sitting next to you who can read incredibly quickly, search through your project, explain unfamiliar code, suggest solutions, generate implementations, and help troubleshoot problems. But you still have to be the one driving the project. You still have to tell that developer what the business actually needs. You still have to review the work. You still have to make decisions. That's much closer to how I think about OpenCode. It's not the person building the business. I'm still responsible for that. It's a tool that helps me build it.

And That's Important for My Clients

This isn't just about making my life easier. That's probably the biggest reason I wanted to write about this. I'm running a small web agency. When I become more efficient, my clients benefit. If I can spend less time fighting development tools, I can spend more time working on websites. If I can move through development tasks faster, projects can move forward faster. If I can investigate problems more efficiently, I can spend more time solving them. If I can experiment with new ideas without an enormous development overhead, I can bring more ideas to the table. That's good for my clients. And ultimately, that's what matters.

I Try to Put Myself in the Client's Shoes

One of the principles I try to follow at Shelby Web Company is simple: I try to treat my clients the way I would want to be treated if I were the client. I try to imagine what it's like to own the business. What would I expect from the person I hired? Would I want someone who simply completed the checklist? Or would I want someone who actually cared about whether the website was helping my business? I'd choose the second one. Every time.

Because a website isn't just a website. For a business owner, it's part of the business. It represents them. It helps customers find them. It answers questions. It generates leads. It establishes credibility. And sometimes it needs to change. That's why I don't want to build websites according to some arbitrary concept of: "Well, we've already spent X amount of time on this, so we're done." If something genuinely needs to be done, then I want to find a way to get it done properly.

Better Tools Don't Mean Less Work

This is another misconception I think is worth addressing. When AI makes development faster, some people assume developers are simply doing less work. I look at it differently. The goal isn't to work less. The goal is to waste less. If a tool eliminates hours of repetitive work, debugging, searching, file navigation, or context switching, that doesn't mean those hours suddenly become meaningless. It means I can redirect that time toward something more valuable. More development. More testing. More ideas. More improvements. More client communication. More strategy. More time thinking about the actual business problem. That's the real promise of AI for me.

OpenCode Is Part of a Much Bigger Shift

OpenCode is just one example of what's happening in software development right now. AI coding agents are moving beyond simple autocomplete. They're becoming increasingly capable of understanding entire projects, working across files, using development tools, running commands, reviewing code, and assisting with larger tasks. OpenCode itself includes different agent modes, including a Build agent for development and a Plan agent designed for analysis and planning. That's a significant shift. We're moving from: "AI, finish this line of code." toward: "AI, help me solve this development problem." Those are two very different things.

What This Means for Small Businesses

You might be wondering why any of this matters if you're not a developer. Here's why I think it matters. The businesses that hire web developers aren't really buying code. They're buying results. They want a website that works. They want customers to find them. They want their business represented professionally. They want problems solved. They want new ideas implemented. And they want someone who will continue helping them as their business changes. If AI-assisted development allows a small web agency to accomplish more of those things without sacrificing quality, that's potentially a huge advantage. It means a small team — or even a solo developer — can have capabilities that previously required significantly more time and resources. And that's exciting.

Why I'm Excited About Where This Is Going

I'm still learning. I'm still experimenting. And I'm sure the tools I use today won't be the exact tools I'm using five years from now. That's okay. I'm not trying to build my business around one particular piece of software. I'm trying to build my business around a principle: Use the best tools available to deliver the best possible work for the businesses I serve. Right now, OpenCode has become an important part of that process for me. It has helped remove a bottleneck. It has given me more flexibility. It has helped me spend more time building and less time fighting my development environment. And most importantly, it has given me another way to create more value for my clients.

The Future of Web Development Isn't About AI vs. Developers

I don't think the future is: AI vs. developers. I think it's: Developers with AI vs. developers without it. The developer still matters. Experience still matters. Business understanding still matters. Creativity still matters. Judgment still matters. But the tools are getting better. And I'm going to keep learning how to use them. Because at the end of the day, that's what being a developer — and a business owner — is all about. Learning. Adapting. Improving. And finding better ways to serve the people who trust you with their business. That's what I'm trying to do at Shelby Web Company. And OpenCode has become one of the tools helping me do it.

Final Thought

Technology shouldn't become another obstacle standing between a business owner and the person trying to help them. It should remove obstacles. That's what I was looking for when I discovered OpenCode. I wasn't looking for a shortcut. I wasn't looking for a way to cut corners. I was looking for a better way to build. And sometimes, finding a better way to build means you can finally stop worrying about the tools... and get back to doing the work.

— Dave Richardson
Founder / Lead Developer
Shelby Web Company
Shelby, North Carolina

Visit Shelby Web Company · Learn more about OpenCode

CallText