Unified Runtime troubleshooting

Fixes for common problems running agents, functions, and API workflows on UiPath Robot 2026.10, such as jobs that stay Pending or never start.

Where to look

Unified Runtime writes to the same logs as the rest of the Robot:

• Robot log: Robot.log, in %ProgramData%\UiPath\Logs\internal in service mode, or %LocalAppData%\UiPath\Logs\internal in user mode. Entries about executor packages and the host process that runs them start with [ExecutorsHost].
• Job logs: the logs of an agent, function, or API workflow job go to Orchestrator and to the local execution log, like the logs of an RPA job.

When the Robot service starts, Robot.log records which executor packages the Executors feed offers, and the workloads each one runs:

[ExecutorsHost] feed '<feed path>' offers <package> <version> [<workloads>]; ...

The robot reports those workloads to Orchestrator. If the log has no such entry, or the Robot couldn't read the Executors feed, the robot reports RPA processes and test cases only.

An agent, function, or API workflow job stays Pending

Cause: No connected robot reports the job's process type. Orchestrator holds the job until a capable robot connects. It doesn't fail.

Solution:

1. Check that the robot runs UiPath Robot 2026.10 (LTS), on Windows, and is connected with client credentials. A robot connected with a machine key doesn't run these workloads.
2. Check that the job's runtime type is Local, or Production (Unattended) for a function or API workflow when the machine template uses Use Production (Unattended).
3. Check that the machine template has Agent slots for agents, or Function slots for functions and API workflows, set above 0. With Use Production (Unattended), check that an Unattended runtime is free.
4. In Robot.log, check for the [ExecutorsHost] feed ... offers entry. If you find "this robot advertises no generic runtime" instead, refer to The robot doesn't report agents, functions, or API workflows.

An agent, function, or API workflow job doesn't start although slots are free

Cause: The tenant has no units left. Agent jobs need units, and function and API workflow jobs need units or an unattended runtime. Without them, these jobs don't start, even when slots are free.

Solution: Check the units available to the tenant, and allocate more. For functions and API workflows, you can also check the unattended runtimes.

The robot doesn't report agents, functions, or API workflows

Cause: The Robot can't read the Executors feed in its installation folder. Robot.log has one of these entries:

• [ExecutorsHost] '<index path>' could not be read (<reason>); this robot advertises no generic runtime.
• [ExecutorsHost] the Executors feed could not be located in the NuGet configuration; this robot advertises no generic runtime and a generic process will not find its executor package.

Solution:

1. Check that the Executors folder exists in the Robot installation folder and isn't empty.
2. Check that the NuGet.config file in the Robot installation folder is present and valid XML. If the Robot can't read it, it can't locate the Executors feed.
3. If either is missing, reinstall the Robot with the UiPath Platform Installer, then restart the Robot service.

An agent, function, or API workflow job fails before it runs

Cause: The host process that runs the executor package didn't start. The job fails, and Robot.log has one of these entries:

• ExecutorsHost exited before signaling HostReady (see [ExecutorsHost] log lines).
• ExecutorsHost did not signal HostReady within <timeout>. The host has 20 minutes to install its packages and start, the same limit as InstallPackageTimeout in UiPath.config.
• ExecutorsHost process exited unexpectedly (code <code>).

Solution:

1. Read the [ExecutorsHost] entries just before the error in Robot.log. They give the reason, such as a package that failed to install. Errors from a proxy or a full disk appear there as the package installation error.
2. Check that the machine has free disk space for the packages folder.
3. If the machine reaches the internet through a proxy, check the Robot proxy settings. Refer to Redirecting robots through a proxy server.
4. Run the job again. After a host process exits, the next job for that executor package starts a new one.