Skip to content

    Type a page, a command such as keryn scan, or a rule id.

    Organization scans

    View as Markdown

    See the AI components and the risk across many repositories in one overview, without cloning anything. Paste several repositories into Keryn Web, or an organisation, user or workspace (github.com/org, bitbucket.org/workspace, @org), and Keryn builds an organization: a saved set of scans with one overview.

    • Public repositories (the organization scan fetches what an anonymous visitor can fetch), or a GitHub token for the requests it needs.
    • A number in mind: anonymous access allows 60 GitHub requests an hour, so a large organisation needs a token. Keryn shows the count before it starts.

    For an organisation, Keryn lists its public repositories and chooses a sensible default set: archived repositories and forks are left out. You can change the selection and filter by language, last push and size. Keryn shows the number of GitHub requests the run needs before it starts, because anonymous access allows 60 an hour; a token raises that to 5,000.

    Repositories go through a queue: a few download at a time and they are scanned one at a time, because Deep is memory-heavy. A rate-limit answer holds the downloads until it resets, and a repository that failed for a reason a retry could fix (a rate limit, a network error) can be retried.

    • Repos by risk, a heatmap, the most common rules, and findings by category with the languages scanned.
    • AI-BOM: the AI components across all repositories, with the agent surface rolled up.
    • A drill-down per repository and per rule, with its scan history.

    Organizations are saved in your browser (IndexedDB), so you can come back and press Scan again. Export an organization report as JSON, Markdown or a printable HTML summary, a zip with one SARIF 2.1.0 file per repository, or the whole organization with its history to import on another device.

    The dashboard of a scanned scope has an Open on trust canvas button. It draws every scanned repository as its own zone, with the model providers and MCP servers it found shared between them, so reliance on one provider across repositories shows (a provider reached from three or more repositories raises an info risk). It is built in the tab from the results already saved there; a repository whose full result is at hand gives its routes and tools, and one held only as a compact scan gives their counts instead of tool nodes. See the Trust Canvas guide.

    An organization is a batch of local scans with a shared index, not a service that holds your repositories. Each repository is downloaded to the tab, scanned there, and kept as a result in IndexedDB on your machine. The only network edge is the fetch of public repositories you named, and the saved set never leaves the browser unless you export it. A repository you scanned once is answered from the rescan cache the next time its commit has not moved.