Introducing Cavendo Engine: Giving AI the Structure Your Business Needs

You Use AI. Now What?

If your team uses ChatGPT or Claude for real work — and most teams do — you already know the problem, even if you haven’t realized it yet.

Someone on your team starts using AI and generates output. A blog post. A competitive analysis. A customer email. A project plan. And then what happens? They paste it into a Google Doc. Email it around. Maybe drop it in Slack. It sits in a chat history that nobody else can see. There’s no review process, no version tracking, no cost tracking, and no way to get that work into the systems your business actually runs on.

We’ve been living in that chaos ourselves. At Cavendo, we’ve spent 25 years helping small and mid-sized businesses build with web technology. When AI became part of the workflow — for us and for our clients — we hit the same wall everyone else did. The output was everywhere. Scattered across conversations, documents, and Slack threads. We were drowning in markdown files with no system at all.

So we built the system we needed.

Today, we’re releasing Cavendo Engine — an open-source workflow platform that brings structure to how teams work with AI.

“We built this by accident. We have been trying to figure out a way to manage multiple revisions and all the back-and-forth with AI agents. It was just quite frustrating. We would have tons of different files, we wouldn’t know what the latest copy was. Then, one of the agents hallucinates and all of a sudden we lost a day’s worth of work. So we said there’s gotta be a better way to have some sort of framework around this output instead of just having it all live in a chat. That’s Cavendo Engine.”

Not Just the Output — The Whole Loop

Most AI tool discussions focus on two things: smarter agents, or easier agent-building. There are frameworks out there — CrewAI, LangChain, and others — that help developers orchestrate complex multi-agent systems. They’re impressive tools, but they’re built for engineering teams at enterprise scale. They solve the question of how do I build a sophisticated AI agent?

That’s not the question most of us are asking.

Across teams, this becomes much more difficult if everybody is just logging into ChatGPT or Claude on their desktop. There is no way for context and project knowledge to be shared. All the more reason you need a structured framework around AI output.

The question most businesses are asking is: how does my team actually work with AI? How do we assign work to AI, give it the context it needs, review what it produces, and get the approved output into our real business systems?

Cavendo Engine answers that question by wrapping a full workflow around AI: both the input and output sides. Here’s how it works:

When a task gets created in Cavendo, the Engine doesn’t just send a bare instruction to an agent. It bundles the relevant project knowledge, past deliverables, revision history, and context into a structured package and pushes that to the assigned agent (whether that’s a Claude API call, a ChatGPT integration, a custom agent, or anything else). The agent gets everything it needs to do the job well, not just “write a blog post” but “write a blog post — here’s the project brief, here’s what was approved last time, here’s the feedback on the last draft, here’s the style guide.”

Then, when the agent produces something, that deliverable flows into a structured review process. Your team (humans, AI reviewers, or automated rules) approves, requests revisions, or rejects the work. Approved deliverables trigger webhooks that route the output wherever it needs to go: your CMS, Slack, email, WordPress, or anything that accepts HTTP.

Here’s what compounds: every review, every approval, every revision becomes context for the next cycle. The agent gets smarter about what your team needs. It’s not starting from scratch every time—it’s building on real feedback.

The full cycle looks like this:

We call this the loop: assign the work, agent executes, submits deliverable, human reviews, routes to your systems, learns from feedback.

“AI is great. It can do a lot very quickly, but you have to give it the right input. You have to give it the background on what it needs to do.”

“So, if you ask it to write a post or write an article, it needs to know everything about the background of that. If it’s for a business, what does the business do? What is their specialty? Who are their customers? What’s their market?”

“If you just load up ChatGPT and say, ‘Write me a blog post based on my website,’ that scratches the surface. It can go out and read the website and pull some ideas from that. But 10 blog posts later, you’ve got a whole history that the AI can continue to build on, learn from, and understand your business and how it operates, and ultimately be able to do its job better, just like an employee would do.”

What’s Under the Hood

Cavendo Engine is open source under the AGPL v3 license, self-hostable, and designed to get out of your way.

git clone https://github.com/Cavendo/Engine.git

Why Open Source? Why AGPL v3?

We’re open-sourcing Cavendo Engine because we believe the foundation should be open. The workflow layer between AI and your business shouldn’t be a proprietary black box that locks you in. You should be able to see what’s running your workflows, how your data is being handled, and what’s under the hood.

We chose the AGPL v3 license specifically. For those unfamiliar, AGPL v3 means the source code is fully available — you can run it, modify it, and self-host it. But if you modify the Engine and offer it as a service to others, you’re required to share those modifications under the same license. This is a deliberate choice. It keeps the project open and ensures that improvements flow back to the community, while preventing the code from being quietly absorbed into proprietary products that compete with the open-source project that made them possible.

For businesses self-hosting Cavendo Engine for their own internal use, AGPL v3 is straightforward — you have the full source, you can customize it, and you’re not obligated to share anything as long as it’s for your own operations. It’s the same license used by projects like MongoDB, Grafana, and Nextcloud — proven infrastructure that businesses rely on every day.

Cavendo Engine is also the foundation we’re building on — for our own clients, for Cavendo Cloud, and for other products down the line. Open-sourcing it means the core is transparent, community-tested, and stable.

“We chose AGPL because we wanted to build something that can genuinely benefit others and that’s sustainable beyond just our team. The code is free. You can use it yourself, use it with your team, build on it. You just can’t wrap it into a proprietary service without contributing back. If that’s something you’re interested in, we’d be happy to chat with you about that — but we believe the foundation should be open. We’ve seen this a lot with AI projects recently, and we feel this is the best direction to go.”

Where We’re Going: Cavendo Cloud

Cavendo Engine is the open-source foundation. Cavendo Cloud is where we’re building the managed experience that most businesses actually need.

Cloud takes everything in the Engine and adds the operational layer: financial dashboards that show AI cost and ROI across your organization, client portals for agencies managing AI work on behalf of clients, daily briefings that summarize what your AI team accomplished, team analytics, and managed hosting so you don’t have to think about infrastructure.

We’re also already running the Engine across our own product portfolio — powering AI workflows in our board management platform, our analytics product, and our creative agency work. We’ll have more to share on that soon, but the short version is: we’re not just building this for others. It’s the infrastructure layer underneath everything we do.

“AI is becoming part of the workforce, and small and mid-sized businesses are next. They need an infrastructure to make that happen. Not some big enterprise system that’s scaled down, but rather something that’s purpose-built for these types of organizations. That’s Cavendo Cloud.”

Cavendo Cloud is coming soon at cavendo.io.

A Different Approach

Most AI tooling is built for two audiences: consumers (chatbots, assistants) or enterprise (orchestration platforms starting at six figures a year). There’s a massive gap in the middle. Small businesses, agencies, mid-market companies use AI every day with zero infrastructure around it.

That gap is where we’ve always operated. And Cavendo Engine is how we’re filling it.

The agent-building frameworks solve a different problem for a different audience. What we’re building is the workflow layer that sits around any AI: the structure that turns scattered AI output into a managed, reviewable, routable part of your business operations.

“Every business is using AI. The ones that figure it out and make it an integral part of their operations — not just a tool people open sometimes — are the ones that are going to pull ahead. That’s what we built Cavendo Engine for, and it’s what Cavendo Cloud will take even further. This technology should be accessible to businesses that don’t have teams of engineers to throw at the problem.”

Get Started

Cavendo Engine is live and open source.

And if you want to be first in line for Cavendo Cloud, sign up at cavendo.io.

Cavendo has been building web technology for businesses for 25 years. Cavendo Engine is our answer to the question every AI-powered team is asking: “We’re using AI — now what?”