Oxelone Docs

Search docs and API

Find a guide, a section of one, or an API operation by its name.

Open console (opens in a new tab)

Connectors Coverage Matrix

Coverage Matrix

The Coverage Matrix answers "Where this workspace is blind, and how blind". It is the page to read before you trust a quiet Alerts list. It is under Collection, beside Connectors.

Who can do it

You want toRole that canPermission
Open the pageOwner, Admin, Analyst, Policy author, Fleet operator, Compliance readermachines:read (View machines)
Create, edit or delete a service category, or add services to oneOwner, Admin, Policy authorpolicies:write (Edit policies)

Without policies:write the Service categories tab draws no controls.

The four tabs

The page has four tabs. Open the one you need: a tab you do not open is not loaded.

Network

The Network tab lists the AI services the agent can recognise, and what each capture strategy can see of them.

Coverage at a glance has four tiles:

TileWhat it shows
WatchedHow many catalogued services are shipped to the fleet. The line under it reads "Every catalogued service ships to the fleet." or "N catalogued but not shipped — switched off, or matching nothing."
DepthThe best case per service across every strategy: content, metadata only or unreachable. "Never summed into one number."
AttributionHow well this build knows each request shape: captured, documented, inferred or unestablished.
Your fleetHow many machines run each strategy: Browser extension, TLS interception, Network steering and CLI & IDE tools.

The four attribution levels mean:

LevelMeaning
CapturedA real request was observed and committed as a fixture.
DocumentedThe vendor publishes the prompt-carrying request shape.
InferredMatchable and blockable by reasoning; content cannot be attributed.
UnestablishedNo hostname could be established. Matches nothing, blocks nothing.

Under the tiles, the page says where it is blind when it is:

  • "N services are a name on a list." "No hostname could be established for them, so they match nothing, block nothing, and attribute nothing. Recorded rather than hidden."
  • "N services live only in mobile apps." "Native iOS and Android traffic is invisible to the extension and the desktop agent alike. No strategy closes this."
  • "No live machine runs ..." with the idle strategies named. "Every cell in those columns is capability, not coverage: what the platform could see if a machine were enrolled to look."
  • "Nothing is enrolled yet." "Every figure above is platform capability. Coverage starts when the first machine enrols on a strategy."

The service list searches by "service, vendor, ID or host" and filters by depth (Content, Metadata, Unreachable), by catalogue (In the catalog, Uncatalogued) and by use (Observed, Not observed). Uncatalogued hosts are ones your fleet reached that the catalogue does not know. If usage cannot be read, the page says "This fleet's usage could not be read, so uncatalogued hosts cannot be listed."

Below the services, the Connections block lists each connection in this workspace and how much its credential can read, such as "N of M record kinds readable", and "Last sync ..." or "Never synced". With none, it reads "No connections yet. Connect a platform under Collection › Connectors and what its credential can read appears here."

Host

The Host tab is the same question asked of files on machines. Host coverage at a glance has five tiles:

TileWhat it shows
ScannersArtifact types this build has a working scanner for.
Seen hereTypes with at least one artifact on this fleet.
Seen onMachines carrying the most widespread type.
Not on older agentsTypes some live agents cannot report.
Agents not reportingAgents that predate the catalogue.

A type with no scanner shows No scanner. A type nothing on your fleet carries shows Not observed. The list searches by "type, category, or tool".

When no agent in this workspace has reported a catalogue yet, the tab says so: "Showing this release's built-in catalogue; no agent in this workspace has reported one yet." or "Showing the newest catalogue reported by agents in other workspaces; none in this workspace has reported yet."

Engines

The Engines tab says, for each kind of thing a policy can look for, whether an engine is running in this workspace.

RowWhat it covers
Personal dataIdentity, contact and financial identifiers in prompts and files.
Credentials & secretsAPI keys, tokens and private keys, checksum-corroborated where the format allows.
Prompt injectionJailbreaks and instruction-smuggling in what reaches a model.
AI supply chainPackages, models and skills a machine pulled, judged against the registry.
Judgment-based policyA judge you author, run against interactions. Asynchronous by nature.
Harmful contentToxic, harassing or unsafe generations in either direction.
Artifact scanningAI-related files on disk, judged by content rather than by name.

Each row ends in a status: Running, No verdict, "Switched off in this workspace." or "Not in this build's registry." Harmful content reads "An engine exists in the tree and is deliberately not registered in this build." Artifact scanning reads "Declared with no runner, so it collects nothing today." If the registry cannot load, the tab reads "Could not load the engine registry, so nothing here can be claimed." See Engines.

Service categories

A service category is a named list of AI services that a Network policy can name in a Service or Host filter as $slug. Oxelone maintains some of them. Your workspace can add its own.

Each category shows its name, an Oxelone or This workspace badge, its description, and counts such as "N services", "N seen here" and "N policies". Open one to see its services, the patterns each matches, and Observed where your fleet has used it.

To manage categories you need policies:write:

  1. Open Collection › Coverage Matrix and choose the Service categories tab.
  2. Choose New category. Give it a Name (the placeholder is "Research tools"), a Description ("AI tools the research team uses"), and choose its Services, or type a site such as chat.example.com.
  3. To change a category, use its menu: Edit… or Delete…. Add services adds to a list; Remove takes out only what your workspace added.

The list searches by "category or service". With none, the tab reads "This workspace has no service categories." With no match it reads "Nothing matches this search."

Every change reaches machines on their next heartbeat.

When it goes wrong

What you seeWhat it means
Could not load the engine registry, so nothing here can be claimed.The Engines tab could not read the registry.
This fleet's usage could not be read, so uncatalogued hosts cannot be listed.The Network list is shown without uncatalogued hosts.

Other read failures show as a red note at the top of the tab that failed.

With the API

Not available with an API key. The Coverage Matrix reads the console's own routes.