How to deploy AI agents in production
The same spec you ran locally, moved to your VPC, Kodeus Cloud, or an airgapped network, without swapping the policy engine on the way.
Blog
Blog
Practical notes on taking an agent from a spec on your laptop to a run you can defend. The first one is how to deploy AI agents in production.
Notes on running agents
The blog is for decisions that do not fit on a product page. Product pages say what the runtime does. A post says how to use that when you are about to put an agent somewhere a user can reach.
There is one post so far. It is the deployment sequence: spec, local run, identity, tools, policy, then the boundary you actually want. More will land when there is a concrete gap, not on a schedule.
Related guides
Production ready AI agents, best MCP servers, LangGraph alternatives, CrewAI alternatives, and LangGraph vs CrewAI.
What you will not find here
No roundups of tools we have not run. No metrics we cannot point at. No claim that you can bring a finished agent from another framework and have us host it. Kodeus drafts the application and operates it. The posts stay inside that.
If you want the product instead of the notes, start at the platform, then developers for the spec, then enterprise if the run has to live inside your perimeter.
Rather see it than read it
Book a demo and we will trace one turn, including a call the policy refused.
Frequently asked questions
Does Kodeus have a blog?
Yes. Posts are practical notes on running agents, starting with how to deploy AI agents in production.
Who is it for?
People deciding how an agent gets from a working spec to something a real user can depend on.
How often will it update?
When there is something concrete to add. The site does not publish for a calendar.
Where are the product pages?
Platform, developers, enterprise, and the guides linked from this index.
Can I request a topic?
Yes. Use the contact page and say what you are trying to run.
Is the SDK documented here?
The developers page covers the spec, the SDK and the local runtime. The SDK is in private preview.
Do posts replace the docs?
No. A post explains a decision. The product pages describe what the runtime does.
How do I get a traced run shown to me?
Book a demo. We walk through one workflow, including calls a guardrail refused.