---
title: "Bastio — The Enterprise AI Security Platform"
description: "Discover Shadow AI in 14 days, redirect risky prompts into a secure multi-model workspace, and build with guardrails. Four pillars on one EU-built platform."
source: "https://bastio.com/docs"
---
# What is Bastio?

Protect model requests, sensitive data, and autonomous agent actions with one observable AI security platform.

Bastio is an open-source AI security platform for model gateways, custom LLM pipelines, and agentic workflows. It detects prompt injection, jailbreaks, PII, secrets, and dangerous tool arguments before they become incidents.

## [Choose your integration](#choose-your-integration)

[Protect an existing LLM client  
\
Route OpenAI- or Anthropic-compatible traffic through Bastio with minimal application changes.](/docs/gateway)

[Build a custom security pipeline  
\
Call detection, PII, and agent-action controls directly and own the enforcement UX.](/docs/developer-apis)

[Run locally  
\
Start the complete self-hosted stack with Docker Compose and verify your first trace.](/docs/getting-started)

[Inspect every endpoint  
\
Use the OpenAPI contract and interactive Scalar reference on a running Bastio instance.](/docs/api-reference)

## [One engine, three control points](#one-engine-three-control-points)

| Control point  | What Bastio evaluates                         | Typical integration                    |
|----------------|-----------------------------------------------|----------------------------------------|
| Model gateway  | Prompts and model responses                   | Change the provider base URL           |
| Developer APIs | Messages, retrieved context, PII, and outputs | Call a focused REST API or SDK         |
| Agent actions  | Final tool names and arguments                | Authorize immediately before execution |

All surfaces feed Bastio's trace and threat model so security teams can investigate the same event developers enforced in code.

## [Platform capabilities](#platform-capabilities)

- **Policy-driven detection:** profile steps define detector, strategy, and threshold.
- **Reversible PII tokenization:** replace sensitive values with stable placeholders and restore approved output.
- **Agent action inspection:** block dangerous shell, filesystem, database, and path patterns before execution.
- **Operational context:** traces connect latency, provider, tokens, cost, environment, and security findings.
- **Deployment choice:** run the open-source stack yourself or use Bastio Cloud's managed gateway and enterprise controls.

## [Recommended reading order](#recommended-reading-order)

1. [Getting started](/docs/getting-started) — run Bastio and verify a protected request.
2. [Architecture](/docs/architecture) — understand the request path and data stores.
3. [Gateway](/docs/gateway) or [Developer APIs](/docs/developer-apis) — implement your control point.
4. [Threat detection](/docs/threats) — tune profiles and actions.
5. [Observability](/docs/observability) — operate the integration in production.

[Getting started  
\
Run Bastio locally, create a gateway, and send your first protected model request.](/docs/getting-started)

### On this page

[Choose your integration](#choose-your-integration)[One engine, three control points](#one-engine-three-control-points)[Platform capabilities](#platform-capabilities)[Recommended reading order](#recommended-reading-order)