> For the complete documentation index, see [llms.txt](https://developer.harness.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://developer.harness.io/ai-security/new-to-ai-security/ai-security-deployment-support-matrix.md).

# AI Security Deployment Support Matrix

<details>

<summary>Updates (July 2026 to September 2026)</summary>

* *September 2026* — Reorganized the matrix into three deployment types: out-of-band instrumentation, gateways, and reverse proxies/load balancers. Added AI Firewall (Redaction) and AI Firewall (Blocking) as tracked capabilities alongside AI asset discovery. For more information, see [Supported deployments](#supported-deployments).

**Previous Updates**:

* *July 2026* — Added *AWS API Gateway* and *Azure APIM* as a supported deployment setup for AI Security.
* *May 2026* — Added *LiteLLM* as a supported deployment setup for AI Security.

</details>

Harness supports AI Security features across specific deployments. The following support matrix lists which deployments currently support AI Security capabilities, such as AI asset discovery, AI Firewall (Redaction), and AI Firewall (Blocking). It also includes the supported versions and any required configurations for each deployment.

The following table describes the capabilities in this matrix:

| Capability              | Description                                                                                                                        |
| ----------------------- | ---------------------------------------------------------------------------------------------------------------------------------- |
| AI Discovery            | Identifies AI assets in your environment, such as AI APIs and MCP assets, and shows how they interact with your APIs and services. |
| AI Firewall (Redaction) | Removes sensitive data from requests before they are sent to an LLM.                                                               |
| AI Firewall (Blocking)  | Blocks malicious requests to AI endpoints, such as prompt injection attempts.                                                      |
| Deployment Version      | The version of the deployment's agent that supports AI Security.                                                                   |
| TPA Version             | The version of the Traceable Platform Agent (TPA) that supports AI Security.                                                       |
| Configurations          | Any configuration required to enable AI Security for the deployment.                                                               |

### Supported deployments

The following table lists AI Security support for each deployment, grouped by deployment type:

<table><thead><tr><th width="141.78515625">Deployment Type</th><th width="157.57421875">Deployment</th><th width="157.7109375">AI Discovery</th><th width="145.07421875">AI Firewall (Redaction)</th><th width="144.125">AI Firewall (Blocking)</th><th width="168.15234375">Deployment Version</th><th width="148.34375">TPA Version</th><th width="159.16796875">Configurations</th></tr></thead><tbody><tr><td rowspan="4">Out-of-band Instrumentation</td><td>eBPF</td><td>Yes</td><td>No</td><td>No</td><td>1.25.2</td><td>1.65.0</td><td>NA</td></tr><tr><td>Mirroring Agent</td><td>Yes</td><td>No</td><td>No</td><td>1.9.2</td><td>1.65.0</td><td>NA</td></tr><tr><td>LiteLLM AI Gateway</td><td>Yes</td><td>No</td><td>No</td><td>1.25.2</td><td>1.65.0</td><td>NA</td></tr><tr><td>Portkey</td><td>Yes</td><td>No</td><td>No</td><td>1.25.2</td><td>1.65.0</td><td>NA</td></tr><tr><td rowspan="3">Gateways</td><td>Apigee**</td><td>Yes</td><td>No</td><td><em>Partial</em></td><td>1.38.0</td><td>1.65.0</td><td>NA</td></tr><tr><td>Azure APIM**</td><td>Yes</td><td>No</td><td>Yes</td><td>1.68.0</td><td>1.68.0</td><td>NA</td></tr><tr><td>Kong AI Gateway**</td><td>Yes</td><td>Yes</td><td>Yes</td><td>2.2.0</td><td>1.65.0</td><td><code>sse_max_size_bytes = 131072</code></td></tr><tr><td>Reverse Proxies / Load Balancers</td><td>Istio**</td><td>Yes</td><td>No</td><td>Yes</td><td>2.2.0</td><td>1.65.0</td><td><code>ext_proc.sse_capture_config.enabled = true</code></td></tr><tr><td>Inline Language Agent</td><td>Python SDK</td><td>Yes</td><td>No</td><td>No</td><td>2.3.1</td><td>1.65.0</td><td>NA</td></tr></tbody></table>

{% hint style="info" %}

* Deployments marked with \*\* support both AI Discovery and AI Firewall.
* *Partial* indicates limited support. Apigee supports AI Firewall (Blocking) only for non-streaming payloads. It does not support streaming payloads that use Server-Sent Events (SSE), in which the LLM sends its response in chunks.
  {% endhint %}

{% @harness-feedback/feedback %}