Skip to main content
Version: 1.36

AI Agent Fleets - Admin Configuration Guide

This guide covers the complete setup and configuration process for enabling AI Agent Fleets in your Okteto organization.

Prerequisites​

Before configuring AI Agent Fleets, ensure you have:

  • Administrator access to your Okteto organization
  • An API key from either Anthropic or AWS (for Amazon Bedrock)
  • A Git token for repository integration (if not, we'll help you set one up)

Enabling AI Agent Fleets for your organization​

Step 1: Enable AI Agent Fleets​

  • From the Okteto Admin Dashboard, navigate to Admin -> AI Agent Fleets underneath the Settings Section
  • Toggle the AI Agent Fleets switch to enable the feature for your organization

AI Agent Fleets Admin Dashboard

Step 2: Configure Access Control​

Choose who can use AI Agent Fleets in your organization:

Option A: Enable for All Users​

  • Select All users to grant access to everyone in your organization
  • This is recommended for smaller teams or companies already familiar with the product

Option B: Enable for Selected Users​

  • Select Selected users
  • Click the users view link to navigate to the user management tab
  • Select the specific users who should have access to AI Agent Fleets

Note: Users without access will not see the Agents tab in their dashboard.

Step 4: Configure Your LLM Provider​

AI Agent Fleets requires an LLM provider to power the Okteto AI Agents. Choose between Anthropic (direct) or via Amazon Bedrock.

Option A: Anthropic Configuration​

Use this option for direct integration with Anthropic's API.

AI Agent Fleets LLM Provider Configuration

  1. Select Provider: Choose Anthropic from the Provider dropdown
  2. Obtain API Key:
  • Click the "Get your Anthropic API key →" link
  • Sign in to your Anthropic account (or create one)
  • Navigate to API Keys section
  • Create a new API key with appropriate permissions
  • Copy the key
  1. Enter API Key: Click Update and paste your Anthropic API key in the field
  2. Save Configuration: Click Save to save your settings

Option B: Amazon Bedrock Configuration​

Use this option if your organization prefers AWS-managed AI services.

AI Agent Fleets LLM Provider Configuration for Bedrock

  1. Select Provider: Choose Amazon Bedrock from the Provider dropdown
  2. Configure Region:
  • Select the AWS region where you have Bedrock enabled
  • Common regions: us-east-1, us-west-2, eu-west-1
  • Ensure Claude models are available in your selected region
  1. Set Up AWS Credentials:
  1. Enter Credentials: Back in the Okteto Admin Dashboard, add your AWS credentials in the API Key field
  2. Save Configuration: Click Save to complete the configuration

Step 5: Configure LLM Model​

The LLM Model section shows which Claude model is being used:

  • Currently, agents use Claude Sonnet 4, optimized for software development tasks
  • This model provides the best balance of speed and capability for code generation
  • Additional models will be available in future releases
note

Note: The model selection is currently managed by Okteto to ensure optimal performance. Custom model selection will be added in a future update.

Step 6: Set Up Git Integration​

Configure Git integration to enable agents to clone private git repositories:

  1. Generate Token:
  • Click "Generate a new GitHub token →"
  • You'll be redirected to GitHub (or continue to your Git provider)
  • Create a token with these permissions:
    • repo (full repository access)
    • workflow (Update GitHub Action workflows)
    • read:user (read user profile)
  1. Configure Token:
  • Copy the generated token from your git provider
  • Click on Configure (Update if you previously had one configured) and paste it in the Git Token field
  • Click on save

Supported Git Providers:

  • GitHub
  • GitLab
  • Bitbucket (coming soon)

Step 7: Verify Configuration​

After completing setup:

  1. Launch Test Agent:
  • Create a simple test agent to verify full functionality
  • Example: "Create a hello world Python script"
  1. Check User Access:
  • Verify the Agents tab appears for authorized users

Configuration Best Practices​

API Key Management​

  • Rotate keys regularly: Change API keys every 90 days
  • Use separate keys: Don't share keys between environments
  • Monitor usage: Track token consumption and costs
  • Set alerts: Configure notifications for high usage patterns in Anthropic

Access Control​

  • Start small: Begin with selected users during initial rollout
  • Train users: Provide guidelines before granting access
  • Review regularly: Audit user access monthly
  • Document policies: Create clear usage guidelines

Cost Management​

For Anthropic Direct:​

  • Monitor token usage in your Anthropic dashboard
  • Set spending limits if available
  • Consider prepaid credits for predictable costs

For Amazon Bedrock:​

  • Use AWS Cost Explorer to track Bedrock usage
  • Set up AWS Budgets for cost alerts
  • Consider Reserved Capacity for consistent workloads

Troubleshooting​

Common Issues​

Users Can't See Agents Tab

  • Confirm user is in selected users list
  • Have user log out and back in
  • Clear browser cache

Git Integration Not Working

  • Verify token has correct permissions
  • Check token hasn't expired
  • Ensure repository access is configured
  • Ensure you are using one of the supported providers (GitHub or GitLab)

Getting Help​

Security Considerations​

API Key Security​

  • Store API keys that you wish to share with Agents securely using Okteto's Admin Variables
  • Never share API keys in code or documentation
  • Use environment-specific keys (dev, staging, prod)
  • Enable API key access logs where available

Compliance​

  • Review your organization's AI usage policies
  • Ensure compliance with data residency requirements
  • Document AI usage for audit purposes
  • Consider GDPR/CCPA implications for generated code

Next Steps​

After configuration:

Create Usage Guidelines: Document best practices for your team Run Training Sessions: Help users understand effective prompting Plan Rollout: Phase adoption across teams Gather Feedback: Create channels for user feedback