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

# Admin guide

> For security, IT and the platform team: what every agent is allowed to do, how you know, and how to deploy and operate xpander.

Your organization's agents run on xpander, so control applies as the agent acts: every model call and every skill call is checked against the agent's policies, given its credential at that moment, counted against its budget, and recorded under the person who asked.

Security review happens once. Every new agent starts inside the skills, models and policies you approved, and inherits that review.

## What you control

| Control | Where | What it does |
| - | - | - |
| Approved skills | **Platform > Agents > Skills registry**, per-agent Add a skill | Which outside systems and actions agents and MCP clients can reach at all, and who sees each |
| Approvals | Per skill ("These actions need approval"), gated commands, permission mode | Which actions wait for a named person's sign-off, and for how long |
| Identity | Per-skill connections: organization or personal; OIDC pre-authentication where systems allow | Whether the agent acts with its own credential or with the asking person's permissions |
| Credentials | Vault, workspace secrets | Credentials are injected at the moment of use. The agent never holds one |
| Models | Platform > AI governance > Models: providers, keys, which models are on and who may use them | Which models agents may use, frontier or open-weight, and through which provider. Every call goes through the [LLM Gateway](https://docs.xpander.ai/run/llm-gateway.md) |
| Budgets | Platform > AI governance > Budgets, per-agent limits | Spend caps per organization, group, member, API key and agent |
| Access | Manage access, Org-wide sharing | Who may use, edit or share an agent |
| Record | Session Manager, task threads, Platform > AI governance > Logs and Usage | Every task, every skill call, every model call and every decision, under the person who asked |

## Where xpander runs

| Deployment | What runs where | Page |
| - | - | - |
| xpander cloud | Everything in xpander's cloud; agents answer to Slack, Teams, an API or a schedule | [Access control](https://docs.xpander.ai/resources/security-compliance/access-control.md) |
| Hybrid | Execution in your Kubernetes cluster, control plane in xpander's cloud | [Hybrid deployment](https://docs.xpander.ai/self-hosted/hybrid.md) |
| Air-gapped | Every component inside your perimeter, including sign-in. With OpenCode and an open-weight model, nothing leaves the perimeter | [Air-gapped deployment](https://docs.xpander.ai/self-hosted/airgap.md) |

Every connection that leaves your perimeter in a self-hosted install is optional and under your control. The egress table on the air-gapped page lists each one, when it is used, and the setting that turns it off or points it inside.

## In this section

> **[Deploy](https://docs.xpander.ai/self-hosted.md)**
>
> Deployments, prerequisites, install, verify, upgrade, scale.

> **[Integrations](https://docs.xpander.ai/run/integrations/ai-gateway.md)**
>
> AI gateway, AI vendors, MCP gateway, desktop AI clients, Slack workspace, Microsoft Teams tenant, email, identity providers, vault providers, EKS clusters, the data layer, file storage.

> **[Licensing and identity](https://docs.xpander.ai/self-hosted/licensing.md)**
>
> The two-file license, renewal, and single sign-on with your identity provider.

> **[Security and compliance](https://docs.xpander.ai/resources/security-compliance/security-architecture.md)**
>
> Architecture per deployment, certifications, data handling.

> **[Access control](https://docs.xpander.ai/resources/security-compliance/access-control.md)**
>
> Members, roles, API keys and audit.


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