DEVELOPER PLATFORM ROADMAP
The ServiceNow Developer Roadmap: Top Announcements from now[dev]26
There are more ways than ever to build on ServiceNow with more support for testing, delivering, and governing what you build. You can build directly on the platform or work in your own coding environment. You can also use dedicated tools to create workflows, AI agents, and user experiences. Build without AI, use AI for specific tasks, or let it carry out defined work for you to review. Whichever path you choose, we’re connecting your work to the data, workflows, and controls it needs for production.
Our first now[dev]26 event brought builders together to explore what’s possible and what’s next. Here are ten highlights from the roadmap session. For each one, you’ll see what it helps you do, what you can use now, and what’s coming next.
HOW TO READ THIS ROADMAP
- Use now: means the capability is available today (Generally Available) subject to eligibility and prerequisites.
- Evaluate next: means the capability is in Early Availability or Controlled Availability. It’s open to selected participants only, and access isn’t automatic.
- Plan around: means the capability is coming next. We include target dates, but timing is subject to change.
────────────────────────────────────────────────────
1. Hand off defined work and spend less time supervising it
If you’re guiding AI through every step, you’re still doing the coordination. We’re giving you a way to hand off defined work while you focus elsewhere. Autonomous Engineer is now in early availability. It uses Build Agent and your team’s development rules to build and test from a plan you review. You still own the requirements, key decisions, and final review. You just don’t have to direct every step.
For work you still want to guide step by step, we’re also improving Build Agent. Soon, it will keep working on longer tasks in the background and let you know when they’re done, so you can move on to something else. It will remember your app’s context and earlier decisions from one session to the next, so you don’t have to explain them again. And with scheduled prompts, you can set routine tasks to run on a regular schedule.
Availability:
- Evaluate next: Autonomous Engineer
- Plan around: Build Agent background tasks and persistent memory (Q1 2027 target)
Learn More:
- For more details on autonomous engineer, prerequisites, and how to get started: Go Here
────────────────────────────────────────────────────
2. Work with source and metadata in one place
Some developers think in records. Others think in code. You shouldn’t need a separate tool for each. ServiceNow Studio now brings both together. The ServiceNow IDE is part of Studio, so you can browse application records or open the Fluent source in the same place. There are no separate tool and no mode to switch into.
You can also manage changes in your way. With Source Control 2.0, you connect to a Git repository and work with it right from Studio. You can even ask Build Agent to push your app to Git for you. Prefer the classic approach? Update sets and application pipelines still work the same.
Two more updates are coming soon. We will introduce Projects to keep everything for one piece of work in one place, including conversations, plans, and tasks. When you open a Project, you’ll see only what belongs to it. Additionally, live preview is also expanding. You’ll soon be able to click part of your app in the preview and tell Build Agent what to change and how.
Availability:
- Use now: Source and metadata in ServiceNow Studio and Source Control 2.0
- Plan around: Projects (early 2027 target); expanded live preview (early 2027 target)
Learn More:
- For more details on ServiceNow IDE prerequisites, and how to get started: Go Here
────────────────────────────────────────────────────
3. Keep tests useful as the application changes
Generating a test is only the start. Keeping it useful as the application changes is the ongoing job. Test Agent is available now in Build Agent in ServiceNow Studio. It helps you generate, run, maintain, and investigate supported functional and UI tests. It builds on the Automated Test Framework (ATF) rather than replacing it.
As your app changes, Test Agent can update the tests that are affected and remove ones you no longer need. When a test fails, it helps you narrow down the cause. You review the findings, fix the issue, and rerun the test. Your generated tests stay available for future regression runs.
Coming soon, Test Agent will run tests in the background so you can keep working. You’ll also be able to investigate failures from the command line. Later, we plan to add code coverage guidance to help you find testing gaps.
Availability:
- Use now: Supported test generation, execution, maintenance, and failure investigation
- Plan around: Background testing (February 2027 target); command-line troubleshooting (December 2026 target); guidance on code coverage (timing to be confirmed)
Learn More:
- For more details on test agent prerequisites, and how to get started: Go Here
────────────────────────────────────────────────────
4. Find out why a flow failed and pick up where it stopped
When a flow fails, the hard part is often figuring out what changed and what went wrong. Starting from the beginning makes it worse. Workflow Studio now gives you more ways to troubleshoot. Flow History lets you compare saved versions of a flow side by side. Flow Execution Analysis summarizes a run, points to likely errors, and suggests the next steps.
Once you know the cause, you can fix it without starting over. In the Flow Debugger, you can change a step’s inputs and resume the run. If a flow errors out, you can resume it from the point where it failed.
Coming soon, you’ll be able to troubleshoot issues directly in a conversation to investigate failures across flows, actions, subflows, and Playbooks. You’ll also get guidance on what to do next. Additionally, we will also be adding the ability to rerun a completed or canceled Playbook. Today, if a guided process ends early or needs another pass, you often must start it over and re-enter what was already captured. With Playbook rerun, you can start it again using the context from the earlier run. That saves time and helps avoid errors from re-entering details by hand.
Availability:
- Use now: Flow History, Flow Execution Analysis, Flow Debugger
- Plan around: Conversational workflow troubleshooting and Playbook rerun (February 2027 target)
Learn More:
- For more details on Workflow Studio supported flows how to get started: Go Here
────────────────────────────────────────────────────
5. Develop in customer context without disrupting everyone else
You shouldn’t have to choose between developing with the context you need and staying out of another developer’s way. Developer Sandboxes give each developer an isolated space on top of a non-production instance. You work with the project’s real apps and configuration. Your changes stay separate from the base instance and from other developers. Sandboxes support parallel work. They don’t replace your shared test and release environments.
Sandboxes come from a ready-made pool, so they’re fast to set up. You can also create and retire them with APIs. Administrators manage sandboxes in App Engine Management Center. Developers can use Build Agent inside their own sandbox. When your work is ready to go live, use source control to review and merge it. Update set syncing moves finished changes between the sandbox and the base instance. You don’t have to export and import them by hand.
Availability:
- Use now: Developer Sandboxes
Learn More:
- For more details on developer sandboxes, prerequisites, and how to get started: Go Here
────────────────────────────────────────────────────
6. Use written approval policy without translating every condition into code
A business policy starts as something people can read. Turning every condition into application logic can leave developers maintaining a second version of the same policy. Available now, Intelligent Approvals evaluates written policy for supported approval decisions. Upload the policy, review its interpreted scope and trigger conditions, and test requests before publishing. Clear cases can be approved or rejected; uncertain cases follow the existing human approval path.
Generated scenarios help developers and policy owners review the interpretation together, identify unclear conditions, and refine the source document. You retain oversight appropriate to the decisions. Coming soon, recommendation mode will suggest outcomes while a human approver makes the final decision. Planned support for APIs and flow actions will enable policy evaluation from code or as a flow step. These are separate extensions, not prerequisites for using policy evaluation today.
Availability:
- Use now: Intelligent Approvals
- Plan around: Recommendation mode (January 2027 target); Intelligent Approvals API and Flow Action (March 2027 target)
Learn More:
- For more details on Intelligent Approvals, prerequisites, and how to get started: Go Here
────────────────────────────────────────────────────
7. Connect outside AI tools and agents to your instance
If you’ve built a useful capability on ServiceNow, you shouldn’t have to rebuild it for every AI tool that needs to use it. Available now, MCP Server Console lets administrators publish supported ServiceNow capabilities as tools for compatible external AI clients. This is inbound access to your instance, not an on-platform agent calling an external tool. Administrators stay in control: they set up the server, choose which tools to publish, and decide which AI clients can connect. Nothing is shared by default.
Your existing security rules still apply. Each tool runs on the platform with the same permissions it has today, so if a user can’t access a record in ServiceNow, they can’t access it through an AI client either. You can open up useful capabilities to AI tools without creating a new way around your access controls. That means an AI tool can work with your ServiceNow data and workflows alongside the files and other systems your team already uses.
Arc is our own example of what that looks like in practice. Now generally available, Arc is a personal, long-running AI desktop agent for enterprise work. Instead of you jumping between ServiceNow, files, and other enterprise systems to finish a task, Arc works across the tools you’ve connected. It plans and carries out supported tasks using memory, reusable skills, and specialized sub-agents. It’s a desktop coordination experience, not another on-platform application builder.
Arc also has guardrails built in. ServiceNow Super Intelligence Control Tower lets your organization govern Arc in one place, while NVIDIA OpenShell limits what Arc can do while it runs. Approval rules decide when Arc can go ahead and when a person needs to step in. Arc can only reach the tools and data you’ve connected and given it access to.
Availability:
- Use now: MCP Server Console, Arc
Learn More:
- For more details on MCP Server Console, prerequisites, and how to get started: Go Here
- For more details on Arc, prerequisites, and how to get started: Go Here
────────────────────────────────────────────────────
8. Build from your own editor with familiar source-driven practices
If your team works in a local editor and reviews code in source control, you can bring those practices to ServiceNow today. The focus here is the guidance and instance context available to coding agents, not a newly launched SDK.
The ServiceNow SDK is the command-line foundation: create, build, install, and retrieve supported application work from your local environment. ServiceNow Fluent represents supported application metadata as TypeScript source, so you can review definitions alongside code and manage changes in source control.
You can write the code yourself or let a coding agent help. Generic coding agents don’t know ServiceNow well on their own. So we give them three tools to fill that gap:
- Build Skills teaches the agent how to build on the platform.
- SDK explain gives the agent documentation that matches the SDK version you have installed.
- SDK query lets the agent look at real records and tables on your instance, so it doesn’t have to guess.
Here’s how it fits together. You write and review code in your own editor. Then you use the SDK to install it on your instance, where the app runs. You keep the workflow you already like, and your coding agent has the ServiceNow knowledge it needs.
Availability:
- Use now: ServiceNow SDK (Build skills, SDK explain, SDK query) and ServiceNow Fluent
Learn More:
- For more details on ServiceNow SDK, prerequisites, and how to get started: Go Here
────────────────────────────────────────────────────
9. Build modern user experiences from your own coding tools
Bringing your own tools and practices to ServiceNow shouldn’t stop at application logic. The same should be true for the experiences people actually see and use. That’s where Lux comes in. Lux is our new UI framework for building user experiences on ServiceNow, and you can build with it in the coding tools you already use. The Lux foundation, Lux Lab, and Lux Lab for VS Code are generally available now
More is on the way. App Shell and Default Record and List open in early availability starting November 5. Component discovery is coming next, and we’re targeting March 2027 for Playbook components for Lux.
Availability:
- Use now: Lux foundation, Lux Lab, and Lux Lab for VS Code
- Evaluate next: App Shell and Default Record and List (Early Availability, starting November 5)
- Plan around: Component discovery; Playbook components for Lux (March 2027 target)
Learn More:
- For more details on Lux, prerequisites, and how to get started: Go Here
────────────────────────────────────────────────────
10. Build assistant tools from work you’ve already documented
Your team has already documented how a lot of work gets done, in catalog items, procedures, and knowledge articles. EmployeeWorks Tool Studio helps you turn that work into assistant tools for employees and then check that those tools get the job done. These tools are separate from Build Agent, which you use to build ServiceNow applications.
Catalog Interpreter turns catalog items into conversations. It keeps the fields and rules you’ve already built. Employees describe what they need, review the details, and submit them. The request then follows your existing approval process. For developers, that means you don’t have to rebuild each catalog item as a separate conversational flow. You maintain one set of rules, and both the form and the conversation stay in sync.
ToolSmith helps you build assistant tools with AI. Soon, it will draft a tool straight from a written procedure or knowledge article. It asks you only about the gaps, so you don’t have to explain the whole process again. For developers, that means less time turning documents into requirements and more time on the logic that needs your judgment. Tools also stay closer to how the business already works, which means fewer rounds of rework with process owners.
Coming soon, evaluation suites will test whether a task was actually completed, not just whether the assistant said it was. This matters because an assistant can sound confident and still file the wrong request or skip a step. Catching that before employees do helps build trust in the tools you release. You define what success looks like. Then you can rerun the tests after every change to catch problems early.
Availability:
- Evaluate next: ToolSmith plugin building, editing, and debugging
- Plan around: Catalog Interpreter (late November 2026 target, starting with Controlled Availability); evaluation suites
Learn More:
- For more details on ToolSmith, prerequisites, and how to get started: Go Here
────────────────────────────────────────────────────
Stay up to date on what’s new and what’s next
The roadmap doesn’t end with this blog. In the ServiceNow Developer Community, you can follow new releases, product announcements, and updates on what’s coming next, find practical guidance, and learn from other developers as you explore new capabilities.
Join the ServiceNow Developer Community to stay connected and keep up with what’s new.
All capabilities mentioned in this blog span current functionality, early availability, and planned investments; they are not all newly launched features. Supported scope, access, prerequisites, and target dates vary. Roadmap plans are subject to change.
