What does iterating quickly mean in the age of AI? For me it means ensuring that my coding agent has all the tools and workflows necessary to ensure that the work it has done is correct, and if its not be told what went wrong.
When building components for Agent Mesh, I like prototyping a use case on the Desktop App and having it run end to end before deploying to my remote instance. This is great as feedback is nearly instant and it allows me to iterate quickly with my coding agent.
The setup for this is simple , first download the Solace Agent Mesh Desktop App and follow the instructions here so that you have the sam CLI on your path. Once you have that, install the AI assistance skills that ship with the CLI, this teaches your favourite coding agent how to build Agent Mesh components. Simply run sam ai-assistance skill install
One cool feature of the CLI is that I can easily apply the declarative config my coding agent wrote to the desktop instance by using --target desktop or if you are using a manifest file then use target: name: desktop(with the proper YAML indentation!). Don’t worry if you don’t know what I am talking about here, once you have the skills setup your agent can simply do this on its own. Just ask it! I found this loop is particularly useful when prototyping custom tools which might require more tweaking in the business logic.
Once you are happy with what you have built against the Desktop App you can easily promote the same config to a remote instance. First export everything you built sam config pull -o ./pulled --target desktop . Then you can use the sam config plan/apply commands against your remote instance to push your changes. By default the apply command is additive only, I usually pass in --prune for a proper overwrite to update my remote instance to match the Desktop App.
If you are curious and want to read more check out these links:
- Declarative config in Agent Mesh: Managing Configuration as Code
- Exporting: Exporting and Migrating
- AI Assistance skills : Building with an AI Coding Assistant (Early Access)