In our previous announcement here on the forum, you’ve seen how coding agents are great at helping you build UiPath automations. But building is only part of the job. Once automations are in production, the time spent operating them and troubleshooting failures compounds quickly.
Today we’re opening public previews for two new capabilities in UiPath for Coding Agents:
- Operate — your coding agent works the same surfaces a human operator works in Orchestrator: jobs, folders, machines, runtimes, resources, access.
- Troubleshoot — your coding agent turns into an investigation agent: describe the problem, it gathers evidence across logs, queues, and traces, tests hypotheses, and tells you the root cause with a concrete fix.
You stay in control at every stage. Actions run under your identity, scoped to your roles. If you don’t have access to certain logs or resources, neither will your coding agent.
You can join the preview on the UiPath Insider Portal here – or keep reading to see what’s new.
Operate with UiPath for Coding Agents: what’s new
Operating with UiPath for Coding Agents mirrors what a human operator works with day-to-day in Orchestrator. Each area exposes commands for inspection (what’s there, what just happened), action (start, stop, share, deactivate, edit), and where it makes sense, historical context (job history, audit logs, Insights aggregations). Today’s release covers:
- Job lifecycle: list, start, stop, restart, resume, history
- Folder and machine inspection: who’s mapped where, which runtimes are available
- Resources: assets, queues, queue items, storage buckets, libraries, triggers, webhooks, calendars
- Licenses and runtime capacity
- Audit logs, credential stores, sessions
- Insights job aggregations: top failures, timelines, per-process performance
- Robot logs, execution traces, and healing-agent diagnostics
More operational activities related to administration and governance are coming soon.
Note: This is a newly built, TypeScript-based CLI, designed from the ground up to be used by both people and coding agents. This is not to be confused with other CLIs UiPath has provided over time (i.e. uipcli).
Example: You wake up to a notification – invoice-processing job failed overnight. You ask your coding agent: Why did the invoice-processing job fail? How can I fix it?
The agent lists faulted jobs from the last few hours, reads the error logs, identifies that failures started after a deployment, checks release notes and versions, and proposes rollback as a safe mitigation. One more prompt and operations are re-established. The rollback is the quick mitigation — the proper fix can be worked on afterwards, off the critical path of a live incident.
Done manually, this investigation could have easily eaten several hours. With a coding agent, it’s minutes.
Troubleshoot with UiPath for Coding Agents: what’s new
Troubleshooting automation failures is hard because the evidence is scattered — job logs live in one place, queue items in another, traces somewhere else, and the root cause is usually a combination of all three. No single tool surfaces the full picture, and no tool tells you what to do next.
Troubleshooting with UiPath for Coding Agents changes that. It combines domain-specific playbooks (knowledge about error patterns and diagnostic procedures) with live platform tools (querying Orchestrator for jobs, logs, traces, queue items, and more) to diagnose issues in context.
Under the hood, every request follows a structured investigation loop: triage → hypothesize → test → root cause → resolution. The agent has actual access to your platform data — not just logs you paste in manually.
Today’s preview covers failures across:
- RPA — faulted jobs, selector issues, activity-level traces, queue item failures
- Orchestrator — permission errors, deployment failures, config drift
- Maestro — process instance failures, step errors, SLA breaches, cross-domain failure chains
Other products and workflows are being onboarded. You’re welcome to try them, but expect more limited results for now.
Heads up: we’re also working to bring troubleshooting into UiPath Autopilot – no terminal required, no extra install, just the chat panel you already have. Stay tuned.
Important: Troubleshooting with UiPath for Coding Agents is a conversational experience
Troubleshooting with UiPath for Coding Agents is designed to work as an interactive conversation, not only a one-shot command. This is the most important thing to understand when testing.
Do not stop after the first response. The agent may need several rounds of interaction to fully diagnose your issue:
- It may ask clarifying questions — answer them to help narrow down the problem (which folder, which process, which environment).
- It may present initial hypotheses — if the first recommendation isn’t right, say so. “That’s not it”, “I already checked that”, “dig deeper”.
- It may need additional context — source files, configs, or details it can’t fetch automatically.
- It can test multiple hypotheses — systematically eliminating causes and deepening the investigation.
Think of it as working with a colleague who’s helping you debug. The more context you provide and the more you engage, the better the results.
Tips for effective troubleshooting:
- Be specific — include job IDs, process names, error messages, queue names.
- Keep the conversation going — if a suggestion isn’t helpful, say so and ask it to continue.
- Redirect when needed — “the issue isn’t permissions, the job ran fine yesterday”.
- Share files when asked — XAML, project.json, configs significantly improve diagnosis.
How coding agents know what to do
A few things make this work across Build, and now both Operate and Troubleshoot:
- The CLI has a unified command structure. Consistency helps coding agents navigate quickly.
- Every command documents itself through --help with consistent usage info, so an agent can introspect anything it isn’t sure about.
- The CLI uses --output json everywhere, so the agent reads structured data instead of trying to parse a UI.
- Skills close the last gap. A skill is a set of natural-language instructions that tells a coding agent which commands to use for a given task, plus any context it might need. Skills install with a single command — uip skills install — and their content is public at github.com/UiPath/skills.
As you keep using a coding agent against your tenant, it picks up the specifics of your environment: naming conventions, folder structures, and so on.
Quickstart
You only need to do this once per machine. After setup, both Operate and Troubleshoot capabilities are always available.
Prerequisite: Node.js (LTS) — it includes npm.
npm install -g @uipath/cli
uip skills install
uip login
…then start your coding agent of choice.
Tell us what you think
You can join the preview on the UiPath Insider Portal here. This is the first cut, and we’d love to hear how it lands with you. If something works well, breaks, or just feels off - let us know. It’s as simple as:
uip feedback send
Your feedback reaches directly the product team. Thank you!