Skip to main content

Agents & Tools

Agents extend the AI's capabilities by giving it access to tools — like searching the web, analyzing data, or running code. When an agent is enabled, the model decides when and how to use its tools to give you better answers.

Built-in agents​

Kognar Platform comes with the following agents pre-installed:

Lets the AI search the internet for up-to-date information.

  • Tools: web_search, web_fetch
  • Use when: You need recent news, documentation, prices, research, or any information that may have changed recently
  • Configuration: Supports Tavily API key for enhanced results; falls back to DuckDuckGo otherwise

Code Review​

Analyzes code for quality, security, and performance improvements.

  • Tools: analyze_code
  • Use when: You want a structured review of a code file or snippet

Writer​

Assists with drafting, editing, and formatting written content.

  • Tools: format_document
  • Use when: Writing professional emails, reports, blog posts, or other documents

Quick Answer​

Optimized for short, direct responses to simple questions.

  • Use when: You need a fast answer without extended reasoning

Analyst​

Analyzes structured data and generates visualizations.

  • Tools: text_analysis, parse_csv, calculate, render_chart, render_table
  • Use when: Working with data, spreadsheets, or metrics that need visual output

Enabling and disabling agents​

Agents can be toggled on or off per conversation:

  1. Click the gear icon in the top-right corner of the chat header to open the right panel
  2. Go to the Agents tab
  3. Toggle each agent on or off

Alternatively, open Settings > Agents & MCP > Agents for a full management view.

When agents are active, their names appear as badges in the input bar.

How agents work during a conversation​

When you send a message with agents enabled:

  1. The model reads your message and decides which tools (if any) to use
  2. Tools are called automatically — you'll see a tool card appear in the response
  3. Tool results are fed back to the model
  4. The model continues its response using the tool output
  5. Multiple rounds of tool use are supported in a single response

Tool cards in the chat show the tool name, status (running / complete), and the result. They are collapsible.

Custom agents​

You can create your own agents with custom behavior:

  1. Go to Settings > Agents & MCP > Agents
  2. Click New Agent
  3. Fill in:
    • Name — Display name
    • Description — What this agent does
    • System Prompt — Instructions for the model when this agent is active
    • Temperature — Response creativity (0 = precise, 1 = creative)
    • Max Tokens — Maximum response length
    • Tool Definitions — Custom tools this agent can invoke

Custom agents appear in the same list as built-in agents and can be toggled on/off.

Importing and exporting agents​

You can share agent configurations across devices or with others:

  • Export: Settings > Agents & MCP > Agents > Export (downloads a .json file)
  • Import: Settings > Agents & MCP > Agents > Import (accepts a .json file)

Rules​

Rules are custom instructions that apply to all conversations. They influence how the model behaves without being a separate agent.

Manage rules in Settings > Agents & MCP > Rules:

  • Name — Label for the rule
  • Type — System (injected into system prompt), Behavior (modifies response style), or Context (adds background context)
  • Text — The instruction text
  • Enable/Disable — Toggle rules without deleting them

Rules are applied globally across all conversations when enabled.