> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://developers.alephant.io/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://developers.alephant.io/_mcp/server.

# AIvis Security Overview

> Understand the AIvis security model, deployment boundaries, and shared responsibilities.

AIvis security starts from a self-hosted enterprise deployment model: the public entry point routes requests into controlled application services, while model services, data stores, workers, and agent execution stay inside the deployment network.

## Security principles

| Principle                              | AIvis behavior                                                                                                                                     |
| -------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------- |
| Private-by-default services            | Only the request router should be exposed publicly. Application, data, model, and agent services should remain on private networks.                |
| Administrator-controlled configuration | Model credentials, connector credentials, user roles, groups, and integration settings are managed from the AIvis administration surface.          |
| Scoped knowledge access                | Connector visibility is controlled by AIvis document access settings, groups, and deployment policy.                                               |
| Separate execution surfaces            | Background indexing, live query traffic, and agent execution run in separate services so that each path can be monitored and scaled independently. |
| Replaceable infrastructure             | Object storage, cache, database, search, and routing components can be mapped to managed services when the deployment requires it.                 |

## Security control domains

| Domain               | What AIvis controls                                                                         | What administrators should review                                                                           |
| -------------------- | ------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------- |
| Identity and session | User sessions, administrator role checks, workspace context, and API authorization.         | Identity provider setup, session lifetime, administrator membership, and account recovery process.          |
| Knowledge ingestion  | Connector credentials, document sets, indexing jobs, and document visibility settings.      | Source-system permissions, connector scope, group mapping, and whether indexed content may be shared.       |
| Model access         | Model credentials, model availability, custom inference providers, and generation settings. | Provider contracts, network reachability, credential rotation, and model usage policy.                      |
| Agent execution      | Agent configuration, approved actions, runtime context, and tool access boundaries.         | Which Agents may call tools, what data each Agent can use, and how failures are logged.                     |
| Operations           | Background jobs, logs, health checks, backups, and deployment configuration.                | Host hardening, network segmentation, incident response, backup restore tests, and observability retention. |

## Data categories

AIvis deployments may process several kinds of data:

* User identity and workspace membership.
* Chat, Agent, and administration activity.
* Connector metadata, document text, chunks, and search indexes.
* Uploaded files and connector-synchronized objects.
* Model provider configuration and credential references.
* Logs, traces, request IDs, and job status.

Treat connector content and uploaded files as sensitive by default. Once content is indexed, access is governed by the AIvis configuration for that connector, document set, group, and Agent path.

## Responsibility boundary

AIvis provides the product controls and service boundaries described in these pages. The deployment owner is still responsible for identity provider configuration, TLS termination, network segmentation, host hardening, backups, secret rotation, observability, and regulatory controls specific to the environment.

## Production readiness checklist

Before broad rollout, verify:

| Check          | Expected state                                                                             |
| -------------- | ------------------------------------------------------------------------------------------ |
| Public entry   | Only the router is internet-facing; internal services are private.                         |
| Administrators | Admin membership is reviewed and tied to the correct organization process.                 |
| Connectors     | Each connector has an approved owner, credential scope, and document visibility model.     |
| Models         | Model credentials are stored through approved secret handling and tested after rotation.   |
| Agents         | Each Agent has an owner, allowed tools, permitted knowledge scope, and rollback path.      |
| Logs           | Logs are searchable by request ID without exposing raw secrets.                            |
| Backups        | Postgres, OpenSearch, and object storage backups have been restored in a test environment. |

## Pages in this section

* [System Description](/aivis-security/architecture/system-description) explains the main services and data stores.
* [Agent Runtime Architecture](/aivis-security/architecture/agent-runtime-architecture) explains how AIvis controls agent execution and tool access.
* [Agent Security Strategy](/aivis-security/architecture/agent-security-strategy) explains how AIvis protects confidential data during Agent execution.
* [Data Flows](/aivis-security/architecture/data-flows) covers indexing, chat, search, and Agent action flows.
* [Data Storage](/aivis-security/architecture/data-storage) describes where AIvis persists data.
* [Access Controls](/aivis-security/architecture/access-controls) covers identity, roles, groups, documents, models, and Agents.
* [FAQ](/aivis-security/architecture/faq) answers common security questions.
* [Deployment Hardening](/aivis-security/deployment-hardening) lists practical controls for production deployments.

AIvis Security documents product and deployment architecture. It is separate from Alephant Gateway policy pages, which describe API gateway routing, prompt, cache, session, and model-provider security.