Something not working?
Most problems have a specific cause and a specific answer. Work down this page in order — the first three checks resolve the great majority of them, and each one takes under a minute.
Check these first
Is it already running?
PathMRI is a single instance. Starting a second copy finds the first one and opens your browser on it. If nothing appears, look for a PathMRI process already listening on port 5180.
Did it fail to start?
The console has no terminal window, so a failure to start is written to
startup-error.log in the data folder — and shown in a dialog. That file names the real
cause.
Does the result say unmeasured?
Unmeasured is not closed. It means the probe could not run on the source host, so the answer is unknown — check that host's credentials and transport before anything else.
Where things live
| What | Windows | Linux / macOS |
|---|---|---|
| Database and logs | %APPDATA%\PathMRI\ | ~/.config/pathmri/ |
| Start-up failures | %APPDATA%\PathMRI\startup-error.log | ~/.config/pathmri/startup-error.log |
| Listen address | appsettings.json | appsettings.json |
| Locked out of admin | PathMRI.exe --reset-admin | PathMRI --reset-admin |
Common situations
A result you do not believe
Open the path in the analyzer and read the two halves separately: what the policy predicted, and what the probe actually saw. A disagreement between them is a real finding, not an error.
If the probe could not run at all, PathMRI says so rather than reporting "closed" — an unmeasured path is never presented as a measured one.
A device will not read
Check the credential and the transport for that vendor on the firewall support page — several devices fall back from an API to SSH, and the fallback needs its own account.
Trust failures fail closed on purpose: if a saved fingerprint cannot be read, the connection is refused rather than trusting a new one silently.
Kubernetes checks are empty
The full engine needs SSH to a node with kubectl, crictl,
jq, ss and dig. A kubeconfig-only connection gives inventory
and drift, and nothing that requires entering a pod.
An upgrade will not install
Close PathMRI first. The installer cannot close it automatically, because the application has no window for Windows to send a close request to.
Reporting a problem
A report with these five things in it can usually be answered without a single follow-up question. A report without them usually cannot be answered at all.
| Include | Why it matters |
|---|---|
| The version | Shown in the console footer and returned by /api/health. |
| What you expected, and what happened | Stated separately. "It does not work" cannot be reproduced. |
| The exact path | Source, destination, port and protocol — the four things that identify a check. |
| Which transport | Agent, agentless SSH, WinRM, ephemeral, tunnel or local. They see different things. |
| The log | startup-error.log for a start-up failure; the Activity Log for anything the console did. |