What's New in Now Assist for Creator: Australia GA Release (May 2026 Store Release)
Test Agent, in-app agents, MCP client, semantic search, and a wider Build Agent footprint across the platform.
Table of contents
- Overview
- The road to Australia GA
- Test Agent
- In-app agents, skills, and agentic workflows
- MCP client integration
- Semantic search
- UI validation tool
- New models, contextual launch, and expanded metadata
- Resources for builders
- Getting started
- What's next
Articles hub
Want to see all of our articles related to Now Assist for Creator? Check out the hub.
Overview
Australia General Availability is here, and we couldn't be more excited to announce it alongside our customer and partner conference Knowledge 2026. If you couldn't make it this year, no worries. We have all the key details as they pertain to Now Assist for Creator right here. If you've been tracking Now Assist for Creator since Zurich, you've watched Build Agent move from an IDE-only preview to a Studio-native experience and now into a release that closes the loop between building, testing, and shipping.
May brings the largest set of additions in the Australia cycle. Test Agent generates and runs ATF coverage from chat. Build Agent can now create AI agents, skills, and agentic workflows inside your custom apps. MCP client support brings external tools into the build flow. Semantic search and a UI validation tool round out a release that makes Build Agent more grounded in your instance and more useful at every stage of development.
For those of you reading this from Knowledge 2026, the Vibe Coding Lounge is the fastest way to get hands-on with Build Agent, part of Now Assist for Creator, live on the floor.
The road to Australia GA
For anyone joining the conversation for Australia GA, here's the short version of how we got here:
In March, the Australia Early Availability release brought Build Agent into ServiceNow Studio, added support for global scope, expanded coverage to more than 35 metadata types across 11 domains, and upgraded the underlying models. The Module Generation Agent joined the portfolio, and Catalog Item Generation picked up edit-and-iterate behavior.
In April, the store release brought everything to Zurich Patch 8, added new metadata types (forms, UI actions, Service Portal, scheduled jobs, events), and improved the keyword search that grounds Build Agent in your instance. The Release lifecycle documentation AI agent launched in early availability, automating release notes and update set descriptions.
May brings all of these and some new exciting capabilities into Now Assist for Creator.
Test Agent
Test Agent extends Build Agent into the parts of the development lifecycle that decide whether what you built is actually ready: authoring tests, running them, and triaging what fails.
Why it matters: The biggest customer pain we hear about testing isn't time. It's skills and resources. Most teams don't have someone whose job is writing and maintaining ATF tests, and the people building apps would rather build than test. As Build Agent scales developer output, that gap widens. Test Agent is what keeps the engine running.
What's shipping: Test Agent is an agentic skill of Build Agent that covers the full test lifecycle in one place: author, execute, troubleshoot. Best of all, it's not a separate product, panel, or license. You prompt Build Agent the way you already do, and and the right skill is activated.
A few examples of what that looks like in practice:
- "Write me an ATF test in global scope to verify all mandatory fields on the incident table are filled before submit."
- "Write me ATF tests for all feature permutations of this app."
- "Execute these ATF tests."
- "Troubleshoot this failure."
When a test fails, Test Agent runs ATF Troubleshooting Agent for root cause analysis on what changed, what failed, and why. It either applies a safe fix on its own or surfaces guidance in the chat panel. Build Agent re-runs the suite until it passes. Tests are saved to the standard ATF tables inside your app scope, the same as tests written by hand. They show up in the same lists, run on the same schedules, and can be promoted as regression coverage.
What makes Test Agent special is it is also the first sub-agent in the Now Assist for Creator portfolio. Specialized agents working alongside Build Agent, each with their own skills underneath, is the direction the rest of the portfolio is heading.
The launch of Test Agent shows we're focused on the full development lifecycle, not just the build step. For example, the Release lifecycle documentation AI agent shipped in April, automating release notes and update set descriptions from your configuration changes. Test Agent joins the portfolio with the same focus on the entirety of the software development lifecycle. Catch up on the April release here if you missed it.
What this means for customers: The workflow fits how developers already work but with a dedicated partner for testing.
1. Build the app.
2. Confirm it does what you intended.
3. Ask Build Agent to write tests for all feature permutations.
4. Run them.
5. Let Test Agent flag the issues you didn't anticipate.
5. Promote the app with measurable evidence of coverage, not just developer confidence.
The same flow works for existing customizations on out-of-the-box applications. Point it at global scope and have it secure behavior you've already built.
Two quick notes. Test Agent captures the behavior you confirm is working. It doesn't independently verify your app is correct before it writes tests, the same way a colleague writing regression coverage wouldn't. And test runs take a minute or so, which is the right trade for getting back a real root cause analysis instead of a generic failure log.
Test Agent generating, running, and triaging ATF tests inside the Build Agent chat panel.
Requirements: ATF Test Generator and Cloud Runner app installed, with a cloud user configured. See the ATF Test Generator and Cloud Runner documentation for setup details.
In-app agents, skills, and agentic workflows
In the March post, we teased the idea of Build Agent building agents. That's here.
Why it matters: Customers are already building custom apps on the platform. The next question is almost always, "How do I add an AI experience to this?" Until now, that meant leaving Build Agent and assembling agents and skills by hand, often without the context of the app already in flight.
What's shipping: Build Agent can now turn a business requirement into a fully configured agent, skill, or agentic workflow inside your custom application. It scans your app's tables, roles, business rules, and existing metadata, then generates the agent along with its instructions, tools, ACLs, and supporting scripts. Every agent is auto-registered in AI Control Tower for governance and observability.
What this means for customers: The same conversational flow that builds your app can now build the AI experience layered on top of it. Custom apps with their own Now Assist panel, governed at the platform level, ship in a single session.
Check out the dedicated post covering this capability in depth, including example prompts and the full set of supported tools.
Requirements: Now Assist for App Engine installed. Access to ServiceNow Studio or the ServiceNow IDE. Admin role. Access to agentic workflows, agents, and skills depends on your ServiceNow AI Platform tier. See Getting started for details.
MCP client integration
Why it matters: Real development workflows span more than one tool. Designs live in Figma, project work in Linear, schemas in Prisma. Every copy-paste between those tools and ServiceNow is a small tax on the build flow.
What's shipping: Build Agent now operates as an MCP (Model Context Protocol) client. Connect a supported MCP server once, and the tools and resources from that server become available in the Build Agent chat panel during build and edit tasks.
Supported servers at GA:
- Figma
- Linear
- Prisma
The MCP connection itself is configured in Workflow Data Fabric and inherited from the instance registry. From there, an admin can enable MCP servers in Build Agent settings and connect to a server in a few clicks.
What this means for customers: External tools participate in the build flow without leaving the platform. A designer can hand off a Figma file, and Build Agent can pull the relevant context as it builds the matching UI. A team can reference a Linear ticket without switching tabs.
Note: At GA, MCP connections are available in the ServiceNow IDE only. Studio support is coming. The MCP connection must be configured in Workflow Data Fabric before Build Agent can connect to it. Admin role required.
Semantic search
Why it matters: One of the most common ways AI-assisted development goes sideways is by creating something that already exists. Keyword search is part of the problem. Looking for "incident process management" only finds artifacts that happen to use those exact words.
What's shipping: Build Agent now uses semantic search to locate relevant tables, scripts, business rules, and other instance artifacts based on meaning rather than keyword overlap. It works in two ways:
- Implicit search. Build Agent decides on its own when to search. As you describe what you want to build, it checks for existing files and applications that match your intent, then surfaces them in the plan for your approval. Helps avoid duplicating something that's already on the instance.
- Explicit search. You can ask Build Agent directly to find something. Useful when you want to locate an artifact without making changes.
What this means for customers: Less duplication, more reuse, and a stronger sense that Build Agent is grounded in what's actually on your instance.
UI validation tool
Why it matters: ATF tests are good at server-side and form-level validation. They're less good at catching the kinds of issues you only see when a UI actually renders, such as components that fail to load, broken navigation, or styling that breaks the layout.
What's shipping: Build Agent recognizes when a build includes UI components and runs Playwright-based UI checks on Cloud Runner. Failures are surfaced with diagnostic context directly in the chat panel. You can also trigger UI validation explicitly when you want a check on demand.
What this means for customers: Visual issues get caught at build time, not after the user files a ticket. Combined with Test Agent, the build now produces functional and UI coverage in the same session.
Requirements: ATF Test Generator and Cloud Runner apps installed on the instance.
New models, contextual launch, and expanded metadata
Three updates that make Build Agent feel more native to how creators actually work.
Additional model support
Build Agent now supports Anthropic Claude on AWS Sonnet 4.6 and Azure OpenAI GPT 5.4, in addition to the models introduced in March. You can choose the provider and model that fits your organization's preferences and constraints.
Contextual launch
Build Agent now picks up context from open ServiceNow Studio tabs and component preview screens automatically. Less manual specification, fewer wasted prompts, and more accurate results on the first interaction.
Expanded metadata support
Build Agent picks up new metadata types in this release:
- Flows
- Service Catalog configurations
- Inbound email actions
- Dictionary overrides
- Choice lists
- Condition builder query conditions
- Enhanced Service Portal capabilities
Each addition closes a small gap that used to require switching between Build Agent and a manual configuration step. The story across the Australia cycle is the same: keep more of the build flow inside the conversational interface.
ServiceNow Studio improvements
ServiceNow Studio Live Preview & UI Editing has arrived. It enables agentic creation of widgets, dashboards, UI elements with visual feedback matching some of capabilities found in ServiceNow IDE.
ServiceNow Studio also picks up a set of smaller improvements at GA that make day-to-day work feel cleaner. Conversational checkpoints let you roll back to any progress point in your current chat. The chat panel itself is more polished, and the differentiation between Studio (UI-first, low-code, declarative workflows) and the ServiceNow IDE (code-first, autonomous full-stack development) is clearer so you can pick the right environment for the work in front of you.
For the latest set of Studio improvements, see Accessing Build Agent in ServiceNow Studio and the ServiceNow IDE.
Resources for creators
Two resources worth keeping close as you work with Build Agent at GA.
Build Agent best practices guide
A new guide covering how to get the most out of Build Agent: working with skills and markdown grounding files, structuring prompts that produce predictable output, setting up working agreements that keep the agent on pattern, and the build-test-iterate workflow that consistently delivers results. Useful whether you're new to Build Agent or looking to lift the quality of your existing sessions.
Read it: Build Agent Getting Started and Best Practices Guide
Prompting guide
The fastest way to learn Build Agent prompting is to ask Build Agent. The companion post walks through how to use Build Agent itself as your prompting coach, including example prompts you can drop into your next session.
Read it: The fastest way to learn Build Agent prompting: ask Build Agent.
Getting started
Platform requirements:
- Australia path: Australia GA family release plus the Australia May store release
- Zurich path: Zurich Patch 9 plus the Australia May store release. We've extended the full set of May capabilities to Zurich customers, so you don't have to upgrade families to get Test Agent, in-app agents, MCP, semantic search, and the rest of what's covered in this post.
Additional requirements:
- Now Assist for Creator subscription, or Build Agent (Trial) via the ServiceNow Store
- Now Assist for App Engine installed for in-app agents, skills, and agentic workflows
- ATF Test Generator and Cloud Runner apps installed for Test Agent and the UI validation tool
- GenAI Controller 13.0.0+ for the Release lifecycle documentation AI agent
A note on ServiceNow AI Platform tiers
The Australia release introduces three ServiceNow AI Platform tiers that determine which AI features, generative skills, agentic workflows, and AI agents you have access to:
- Foundation: AI basics that deliver insights
- Advanced: AI to boost productivity across relevant use cases
- Prime: Full autonomous AI, including the ability to create your own AI agents and skills
In particular, generating runtime agents and skills inside custom apps depends on Prime tier access. Confirm with your account team before planning Prime-only workflows on a Foundation contract.
Available now: May 5, 2026 (Australia GA and Australia May store release)
What's next
A look at what we're working on past GA. Directional, subject to change.
Build Agent in implementations. Customers spend significant effort customizing out-of-the-box applications during initial implementations. We're working on the ability for Build Agent to help with this work, starting with ITSM and expanding from there.
Further metadata support. Build Agent's metadata coverage continues to grow. The goal is straightforward: more of what you build on the platform should be reachable from the conversational interface, with less switching out to manual configuration.
More sub-agents. Test Agent is a preview of where the Build Agent skill model is headed. Expect more specialized agents that operate alongside Build Agent for focused parts of the development lifecycle.
Expanded skill capabilities. Building on the in-app agents foundation, we're investing in deeper support for skill creation, including richer tool options and better grounding for skills that work across application boundaries.
MCP in Studio and more servers. Bringing MCP client connections into ServiceNow Studio so the experience is consistent regardless of where you build, and expanding the set of supported MCP servers beyond Figma, Linear, and Prisma.
Alongside new features, we're investing in stability, performance, context handling, and the small details that make Build Agent something you can trust on real work. The feedback you've shared with us through the community, in customer conversations, and during release previews directly shapes that list.
Australia GA is the most complete version of Now Assist for Creator we've shipped. Build Agent is now part of a working flow that goes from idea to app to tested and documented release, all inside the platform.
There's more on the way, and we'd rather show you than tell you.
Try it on something real. Build the app you've been putting off. Add an agent to a custom app you already shipped. Tell us what worked and what didn't. The feedback you share is what shapes the next release.
Try it. Tell us about it.
- On Australia or Zurich Patch 9? Install the Australia May store release and start building.
- Already building? Tell us what you've shipped and what we should improve. Drop a comment below.
- Found this helpful? Mark it "Helpful" and share it with your team.
- Want more? Visit the Now Assist for Creator hub.
Now Assist for Creator: Australia GA + May 2026 Store Release | Available May 5, 2026
