> ## Documentation Index
> Fetch the complete documentation index at: https://docs.xenovia.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Discovery

> One list of the agents, coding agents and AI apps your organization uses, with what is risky about each.

See every agent and AI app your organization uses, and which ones aren't governed yet. Discovery is one list: the agents in your code, the coding agents your developers run, and the AI apps your people sign in to or install.

For example, connect Okta, and an AI writing app 40 people sign in to shows up as **Shadow AI**: used, but nothing governs it. From its entry you can connect an integration that governs it, or block it.

Each entry shows who found it, who uses it, what is wrong with it, and whether Xenovia governs it. Discovery fills from the integrations you connect in **Settings › Integrations**. Admins can open that page from Discovery with **Add integration**.

![Discovery page with Found, Needs review, Shadow AI and Protected counts and a table of apps and agents with their status](https://mintcdn.com/xenovia/RO5UF1Tk8L9B7yPt/images/screens/discovery-overview.png?fit=max\&auto=format\&n=RO5UF1Tk8L9B7yPt\&q=85\&s=63347774910dce491a0615d86a3b540a)

The Discovery page

## Where the entries come from

Each integration adds what it can see. An app that several integrations see is one entry, with every source listed under **Found by**.

| Integration | What it adds | How we know |
| - | - | - |
| [GitHub](https://docs.xenovia.io/discovery/github.md) | Agents, LLM integrations, MCP servers and MCP clients in your repositories | Found in code |
| [Claude Code, Codex, Cursor](https://docs.xenovia.io/agents/coding-agents.md) | The coding agent, and the developers using it | Hooks reported use |
| [Claude Enterprise](https://docs.xenovia.io/agents/ai-platforms.md#claude-enterprise) | Who uses Claude, Claude Code and Cowork in your Claude organization | Seen in the platform's records |
| [Claude Platform](https://docs.xenovia.io/agents/ai-platforms.md#claude-platform) | The Managed Agents in a Claude Console workspace | Seen in the platform's records |
| [OpenAI Platform](https://docs.xenovia.io/agents/ai-platforms.md#openai-platform) | Every API key in your OpenAI organization, with last use and the past week's usage | Seen in the platform's records |
| [Okta, Microsoft Entra ID, Google Workspace](https://docs.xenovia.io/discovery/integrations.md) | AI apps your people sign in to, or have given access to work data | Signed in through single sign-on; Was given access to work data |
| [Jamf, Microsoft Intune](https://docs.xenovia.io/discovery/integrations.md) | AI apps installed on managed devices, and the devices themselves | Installed on a device |

Calls that go through a Xenovia agent add **Went through the Xenovia gateway**. Identity providers and device management see every app, not only AI: Xenovia keeps the apps its AI app catalog recognizes and skips the rest.

## The Discovery page

| Card | What it counts |
| - | - |
| **Found** | Agents, coding agents and AI apps |
| **Needs review** | Nobody approved them yet |
| **Shadow AI** | Used, but not connected |
| **Protected** | Checked before they act |

The table lists each entry with its **Name**, **Source**, **Status** and **Last seen**. Search it, or filter by **Type** (Agents, Coding agents, AI apps), source and status. When GitHub is connected, **Scan now** scans your repositories again.

An entry's status is worked out from what is open against it. The first that applies wins:

| Status | Meaning |
| - | - |
| **Blocked** | An admin reviewed it and chose **Block** |
| **Shadow AI** | Only sign-ins, access grants or installs show it is used, nobody reviewed it, and no integration that can govern it is connected |
| **Hooks missing** | A coding agent sent no hook events in 7 days |
| **Protected** | No warning or critical finding is open, a live Xenovia agent governs it, and its traffic went through Xenovia in the last 30 days |
| **Needs review** | A warning or critical finding is open, or nobody reviewed it and no Xenovia agent governs it |
| **Approved** | None of the above. It shows **Connected** when a Xenovia agent governs it |

## Findings

A finding is one thing wrong with an entry. It is **Reported by** an integration (for example a GitHub scan), comes from **Xenovia's checks**, or was **Found while planning**. Xenovia's checks run over everything Discovery knows after every sync, scan, review, connect or disconnect, and once a day:

| Check | Reports a finding when |
| - | - |
| Shadow AI | Only sign-ins, access grants or installs show the app is used, nobody reviewed it, and no integration that can govern it is connected. Critical from 10 team members, a warning below that |
| Not governed | An integration that can govern the entry is connected, but the entry has no Xenovia agent, or its agent isn't live |
| Hooks missing | A coding agent sent no hook events in 7 days |
| Risky app | The catalog flags the app: it trains on your data (critical), has no SSO, or has no retention controls (warning) |
| Not reviewed | It was found more than 14 days ago, nobody reviewed it, and no Xenovia agent governs it |
| No owner | Nobody owns the app or agent |
| Blocked but still used | It was reviewed as blocked, and someone signed in, granted access, or sent a hook or gateway call in the last 7 days |

A finding is **Open** until the next scan or check no longer reports it; then it is **Resolved**, and it opens again if it comes back. An admin can mark a finding **Accept risk**, with a reason. An accepted finding stays closed until someone chooses **Reopen**.

## From a finding to a governed app

For an agent found in your code, see [GitHub](https://docs.xenovia.io/discovery/github.md#from-code-to-a-governed-agent). For an AI app or platform:

**1. Plan a fix**

Open the entry and choose **Plan a fix**. Xenovia looks at the entry, its open findings and what your connected integrations can do, and proposes one or more fixes: **Connect an integration** that governs the app, **Route it through Xenovia** for an app whose traffic can go through an agent endpoint, or **Block or remove it** through a connected identity provider or device management integration when nothing can govern it.

**2. Apply it**

Choose **Apply** to make a change, or **Discard**. Xenovia checks again on the next scan.

For shadow AI, an admin can also choose **Review**. **Approve**: allowed, and Xenovia stops asking about it. **Block**: not allowed, and Xenovia keeps listing it and flags anyone still using it. This records a decision; it doesn't stop the app. To stop it, plan a fix.

## Entry pages

| Tab | What it shows |
| - | - |
| **Overview** | What it is, its vendor and category, the Xenovia agent that governs it, when it was first seen, and who found it |
| **Findings** | Every finding, open ones first, with its severity, evidence, source and status |
| **Team members** | Who uses it |
| **How we know** | What each integration saw that put it on the list |

A device found by Jamf or Intune has its own page: the person, platform and serial number, and the AI tools installed on it.

## Who can do what

Every member of the organization can read Discovery. Only admins can add integrations, start a scan or a plan, apply a setup or a fix, review an entry, and accept or reopen a finding. Repository scans and plans use AI credits. While your organization has none left, scans and plans don't run.

## What next?

> [card] [Connect GitHub](https://docs.xenovia.io/discovery/github.md)
>
> Find the agents in your code.

> [card] [Add an integration](https://docs.xenovia.io/discovery/integrations.md)
>
> See the AI apps your people use.

> [card] [Protect a coding agent](https://docs.xenovia.io/agents/coding-agents.md)
>
> Check what Claude Code, Codex or Cursor does.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.