AI Skills: Your AI Coding Agent Can Now Build on Solace Without Guessing
New, agent skills ground Claude Code, Cursor, and other coding agents in real Solace docs and real-time data streaming best practices.
AI coding agents are an important tool in the toolbox that developers use and adding Agent Skills in Solace Platform provides them the grounding they need to help you get to production code.
These AI agent skills ground agents in canonical Solace documentation and best-practices before they generate a single line of code. These skills have been tested and validated across a number of scenarios, but like all AI-based coding from “Vibe” through to “AI Augmented Software Engineering”, we do still need to point out that you are the developer and the responsibility for the final code is yours. This first release helps to get you there much faster, and we will continue to add and update skills (and the APIs on which they rely) with every new Solace Platform release.
What the skills do
- Design mode helps you pick the right messaging pattern, publish/subscribe or request/reply, and the right delivery mode, guaranteed or direct.
- Implement mode generates a runnable Maven project wired to your existing broker.
- Topic architecture guidance is read from our canonical best practices page.
- Every generated project gets a verification checklist, separating what the skills handled from what is yours to review.
- A built-in feedback path turns your session into a ready-to-send report for Solace support, our ideas portal, or this community.
Check out a great demo of it in action
What’s in it
- solace-application-development is the umbrella skill and single starting point. It has a Design mode that helps pick the right messaging pattern (guaranteed pub/sub by default, direct pub/sub or request/reply when that’s what the workload calls for) and an Implement mode that generates a runnable Maven JCSMP application, along with a verification checklist tailored to what it just built.
- solace-topic-best-practices reads the live Topic Architecture Best Practices page and applies it to topic-hierarchy decisions, instead of the agent guessing at a topic structure from pattern-matching.
- solace-messaging-feedback turns your session into a copy-paste-ready bug report or feature request, and routes it correctly on its own: a support email if you have a contract, a post to the ideas portal for feature requests, or a Solace Community post if you don’t have a contract yet.
Why it’s worth using
Every named API call, parameter, and config value the skill produces traces back to an actual canonical Solace doc, not something inferred from stale training data. The patterns it defaults to are the ones we already know scale: binary payloads instead of XML, publishing to a topic instead of writing straight to a queue, a queue with a topic subscription instead of a topic endpoint, and delivery-mode guidance that steers you away from settings that look fine in a demo and cause problems later. Every file it generates carries a plain disclaimer that it is AI-generated and should be reviewed before production, so you know exactly what to check before it ships.
Broker setup, TLS configuration, auditing an existing brownfield app, tuning flow control, anything on the management API: this skill is built to recognize those requests and point you to the right place instead of confidently improvising an answer. We tested for that explicitly. An agent that says “that’s outside what I cover” is more trustworthy than one that guesses, and we’d rather ship the former.
Get started
JCSMP is the first API flavor, since it’s our most widely used messaging API, and more are on the way. Install the skill, point your agent at a real Solace messaging task. Check the links in the left menu and get started now.