---
title: "Enabling the MCP Server"
slug: "polarity-mcp-enabling-the-ai-feature"
description: "Learn how to enable the Polarity AI feature in the Admin panel, including model settings configuration for chat and MCP LLM reducers."
tags: ["admin panel", "AI feature", "chat", "enable AI", "LLM", "MCP reducers", "model settings", "server configuration"]
updated: 2026-08-04T16:58:08Z
published: 2026-08-04T16:58:08Z
canonical: "knowledge.threatconnect.com/polarity-mcp-enabling-the-ai-feature"
---

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

# Enabling the MCP Server

**The Polarity AI Features, which power the built-in chat and LLM-optimized reducers for the MCP, is configured in the Admin panel under Server Configuration.**

> [!WARNING]
> **You must be a Polarity administrator to access these settings.**

## Step 1: Enable the MCP Server

1. Log into the Polarity Server’s operating system CLI with a user who has “write” access to the Polarity Server’s install directory (`/app` by default).
2. Open the `.env` file for editing.
  1. In a standard installation, this file will be located at `/app/.env`.
3. Add the following line to the end of the file:

```plaintext
HERMES_MCP_SERVER=true
```
4. Save and close the file.
5. Restart the Polarity Server containers.
  1. `cd /app && sudo ./down.sh && sudo ./up.sh`

## Step 2: Navigate to Model Settings

1. Log in to the Polarity interface.
2. Click the “Cog” (or “Gear”) icon in the upper right-hand corner.
3. Select “Settings” from the drop-down navigation.
4. Select “Server Configuration”.
5. Click the “Model Settings” tab from the navigation on the left.

## Step 3: Enable AI

1. Toggle the AI-enabled switch to the on position.
  1. *This activates the AI backend.*
  2. *Without this toggle, the chat endpoint will return an error, and reducers will not be applied during MCP lookups.*

## Step 4: Select a Provider

1. Use the Provider dropdown to select your AI provider.
  1. *The available providers are described in* [*"****AI Provider Reference****"*](/v1/docs/polarity-mcp-ai-provider-reference)*.*
  2. *Once you select a provider, the Model dropdown will populate with the available models for that provider.*

## Step 5: Enter Provider Credentials

**Depending on the provider, you will see different credential fields:**

| **Provider** | **Required Credentials** |
| --- | --- |
| `openai`, `anthropic`, `gemini` | API Key |
| `bedrock` | AWS Access Key ID, AWS Secret Access Key, AWS Region, and optionally AWS Session Token |
| `ollama` | Base URL (e.g., http://localhost:11434). No API key required. |
| `openai\_compatible` | Base URL (pointing to any OpenAI-compatible API). API Key if required by the endpoint. |

> [!NOTE]
> **All credentials are encrypted at rest using Polarity's Vault cipher.**
> 
> They are *never* stored as plaintext in the database.

## Step 6: Configure Advanced Settings *(Optional)*

**Expand the Advanced Settings section to adjust the following parameters:**

| **Field** | **Default** | **Description** |
| --- | --- | --- |
| Temperature | `0.7` | Controls response randomness. Lower values (0.1-0.3) produce more deterministic, factual responses. For security analysis workflows, values between 0.1 and 0.4 are recommended. |
| Max Tokens | `4096` | The maximum number of tokens the model may generate in a single response. |
| Max Depth | `25` | The maximum number of sequential tool calls the LLM may make in one conversation turn. Limits runaway agentic loops. |
| Timeout | `Server default` | Maximum wait time in milliseconds for a response from the AI provider. Set to 0 to use the server's HTTP idle timeout. |
| Metric Retention Days | `180` | How long conversation metrics and token usage records are retained. |

## Step 7: Test the Configuration

1. Click the Test Chat button.
  1. *Polarity sends a single verification message to the configured AI provider and displays the response.*
  2. *A successful test confirms that the provider, model, and credentials are correctly configured.*

## Step 8: Save

1. Click Save.
  1. *The settings take effect immediately.*
  2. ***No server restart is required.***