Point StackXray at your code. Get a ranked, evidence-backed map of the capabilities worth rebuilding as AI agents, scanned in an isolated sandbox, never stored, never trained on.
Scan it here
This repo is private. Add a read-only access token so we can read it (used once, never stored).
Verify your email to run this scan
We'll send a one-time code — no password needed.
You've used your free token allowance. Bring your own Anthropic API key to keep scanning.
Your code is parsed into a private, searchable index, and the copy you provide is deleted right after indexing.
Add ticket exports (optional, unlocks "Observed demand")
Connect Jira or ServiceNow so each finding shows how many real work items relate to it -- deleted with your code the moment the scan finishes.
Add 30+ days of access logs and we'll show which capabilities saw zero requests — reported as candidates, never as "retire" outright. Deleted with your code the moment the scan finishes.
Free up to ~100k tokens, about one codebase. Larger runs need your own key.
Run it yourself
StackXray's scanner is open source. Download it and run the scan entirely on your own computer, with your own credentials. Nothing is uploaded, ever.
Run it LocallyWhat happens to your code. It is fetched into an isolated sandbox, parsed (never executed), and deleted the moment the scan finishes. We do not keep it, and we never train on it. Only the resulting report is stored, at a private link, and purged after 30 days.
Fetching your code into a sandbox.
This can take a minute or two for larger codebases.
FAQ
Everything you need to know before you point StackXray at your code.
It is fetched into an isolated sandbox, parsed and never executed, then deleted the moment the scan finishes. We do not keep it, and we never train on it. Only the resulting report is stored, at a private link, and purged after 30 days.
A capability is a distinct piece of behavior your product already has, found by reading the code itself, not guessed from a description. Each one is scored by how well an AI agent could take it over today, and ranked by that score against how much of it is duplicated elsewhere in your codebase.
Export your tickets as CSV from either system and upload them, up to 3 files combined. We match ticket volume back to the capabilities we found in your code, so each finding shows how much real, observed demand there is for it, not an estimate. For a public GitHub repo we also read its issues automatically, no export needed.
Pasting a URL works for any public repository. Connecting your GitHub account lets you pick a private repo without handling a token yourself, and StackXray only ever reads it, once, for that one scan.
Yes. StackXray's scanning engine is open source. "Run locally" takes you straight to the repository if you'd rather point it at your code from your own machine, with your own credentials, and never send anything to us at all.
30 or more days of access logs let us flag capabilities that saw zero real requests in that window. Those are reported as candidates to retire, never as a verdict, since low traffic isn't always the same as unused.
Loading your repositories…