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

# Agent Configuration

> Configure AiKA agent behavior — control query processing, tool usage, response generation, and model settings.

Configure AiKA's agent behavior to control how it processes queries, uses tools, and generates responses.

## Overview

AiKA's agent can be configured through the **Portal's Plugins** settings:

**Plugins → AiKA settings → aika → agent**

These settings control the agent's reasoning behavior, tool usage limits, and response generation.

## Configuration Options

### Temperature

**Path**: `aika > agent > temperature`

Controls randomness in the agent's responses.

* **`0`**: Deterministic, focused responses (recommended for most use cases)
* **`0.5`**: Balanced creativity and consistency
* **`1`**: Maximum creativity and variation

**Default**: `0`\
**Range**: `0-1`

### Show Reasoning

**Path**: `aika > agent > showReasoning`

Display the agent's internal reasoning/thinking process to users.

* **`true`**: Shows thinking traces from reasoning models (e.g. GPT-5, Gemini-2.5-pro)
* **`false`**: Only shows final responses

**Default**: `false`

<Note>Non-reasoning models ignore this setting.</Note>

### Reasoning Effort

**Path**: `aika > agent > reasoningEffort`

Controls how much computational effort reasoning models invest in their thought process.

* **`low`**: Faster responses with basic reasoning
* **`medium`**: Balanced reasoning depth and speed
* **`high`**: Deeper reasoning but slower responses

**Default**: `low`

<Note>
  Currently only applies to OpenAI reasoning models (gpt-5.1, gpt-5.2).
</Note>

### Priority Processing

**Path**: `aika > agent > priorityProcessing`

Enable priority processing for OpenAI API requests to get significantly lower and more consistent latency.

* **`true`**: Enables priority tier (premium pricing)
* **`false`**: Standard processing tier

**Default**: `false`

<Note>
  * Only applies to OpenAI models
  * See [OpenAI Priority Processing docs](https://platform.openai.com/docs/guides/priority-processing) for pricing
</Note>

### Max Steps

**Path**: `aika > agent > maxSteps`

Limit the maximum number of steps (tool calls + reasoning iterations) the agent can take before stopping and providing a response.

**Default**: `10`

**Recommended values:**

* **`5`**: Simple queries, tight cost control
* **`10`**: Balanced for most use cases (recommended for production)
* **`20`**: Complex research and multi-step reasoning
* **`undefined`**: No limit (not recommended for production)

#### Why Use maxSteps?

Without a step limit, the agent could:

* Make excessive tool calls, increasing API costs
* Take too long to respond to complex queries
* Get stuck in reasoning loops

With `maxSteps`, the agent will:

* Stop after reaching the limit
* Ensure predictable response times and costs

### Custom Instructions

Add custom instructions to guide the agent's reasoning and response formatting.

#### Reasoning Instructions

**Path**: `aika > agent > reasoningInstructions`

Guide how the agent thinks and makes decisions.

**Example**:

```
Always consider security implications.
Prioritize official documentation over community sources.
Verify information from multiple sources when possible.
```

These instructions are injected into the agent's system prompt after the knowledge integration section.

#### Answer Instructions

**Path**: `aika > agent > answerInstructions`

Guide how the agent formats and delivers responses.

**Example**:

```
Always provide code examples when relevant.
Use bullet points for lists of 3+ items.
Include links to relevant TechDocs pages.
Structure responses with clear sections using headers.
```

These instructions are injected into the response standards section of the system prompt.

## Related Configuration

* [AI Gateway Configuration](../ai-gateway/index) - Configure AI providers and models
* [Getting Started](./getting-started) - Initial AiKA setup
* [MCP in PortalServers](../mcp) - Configure MCP tools and third-party servers