Configure from Slack

Agents can be configured directly from Slack.

Step 1: Access the Agents management panel

From Slack, use the /agent slash command to access the agents management panel:

Click on Start An Agent to open the panel:

Click on Create Agent :

Click on Self-Hosted:

Step 2: Create the Agent

The agent requires a few parameters to operate properly.

• Agent Name - a unique, meaningful name for the agent

• Runner - the Local runner which will host the agent. If you haven't created a local runner yet - please refer to the Local Runners Documentation first

• Agent Description - A brief description which contains information on what the agent is meant for

• Docker image - The Docker image we have built earlier, based on the kubiya/base-agent:latest image. You can use the original base agent image if you do not need to extend the base image

  
• Model (LLM) - Agents can use different large language models to operate. The default option is gpt-4 but you can define your agent with another model based on the specific purposes it requires to fulfill. Following are the supported models

  • You can find more details about the models at the following link.

• Access - define RBAC on which users and groups can access this agent. Users who are not defined can later create an access request.
• Agent Instructions - this prompt is used to define what the agents do and provide it with detailed instructions on how to operate and how use cases can be consumed using it

• Integrations - connectors to attach to this agent. Will require any user who attempts to run the agent to connect with the relevant integrations (eg. connect GitHub with Kubiya)

• Secrets - secrets to decrypt and pass to the agent container as environment variables using runtime - useful for API keys, secrets, etc

  
• Environment Variables - providing a flexible way to configure and customize the behavior of the tools running inside the container.

  
• Additional Knowledge - Links to Github gists that contain additional instructions on top of the agent's instructions. This enables the extension of the context in which the agent provides.

  

  

  Gist with an example of content
• Click on Create to create a new agent, and fill in the relevant details : Awesome! You can now access your agent