Skip to main content

Quick Start Guide

Complete your first Outermind workflow by setting up email monitoring for a mailbox.

Overview​

This guide walks you through the essential steps to get Outermind processing emails. By the end, you'll have:

  1. Configured an LLM provider
  2. Created a specialist agent
  3. Added a monitored mailbox
  4. Verified email processing works

Estimated time: 15-20 minutes

Prerequisites​

Before you begin, ensure you have:

  • Admin access to Control Bridge
  • An API key from an LLM provider (Anthropic, OpenAI, or Azure OpenAI)
  • Access to a Microsoft 365 shared mailbox

Step 1: Configure an LLM Provider​

First, you need to connect Outermind to an AI service.

Add a Provider​

  1. Navigate to Configuration > AI Providers > LLM Providers
  2. Click Add Provider
  3. Select your provider type:
    • Anthropic - For Claude models
    • OpenAI - For GPT models
    • Azure OpenAI - For Azure-hosted models
  4. Enter your API credentials:
    • Name - A friendly name (e.g., "Production Anthropic")
    • API Key - Your provider's API key
    • For Azure, also provide the endpoint URL
  5. Click Save

Verify Connection​

After saving:

  • The provider should show a green "Active" status
  • If it shows an error, double-check your API key

Tip: Start with Anthropic Claude Sonnet for the best balance of capability and cost.

Step 2: Create a Specialist Agent​

Agents are the AI workers that process emails. Create your first one:

  1. Go to Configuration > AI Agents > Agents
  2. Click Add Agent

Configure Basic Settings​

  1. Name - Give it a descriptive name (e.g., "Customer Support Agent")
  2. Description - Briefly describe what this agent handles
  3. Model - Select the LLM model to use (e.g., "claude-sonnet-4-6")
  4. Temperature - Keep at 0.7 for balanced responses

Write Instructions​

The instructions (system prompt) define how your agent behaves. Here's a template:

You are a customer support specialist for [Company Name]. Your role is to:

1. Respond helpfully and professionally to customer inquiries
2. Look up relevant information using available tools
3. Escalate complex issues that require human intervention

Guidelines:
- Be friendly but professional
- Keep responses concise
- If unsure, ask clarifying questions
- Never share sensitive internal information

Tip: Click "Improve with AI" to get suggestions for better instructions.

Assign Tools​

Select the tools this agent can use:

  • reply_to_email - Send email responses (essential)
  • search_knowledge_base - Look up information
  • request_human_input - Request human input or approval
  • lookup_employee - Find employee information

Click Save when done.

Step 3: Add a Monitored Mailbox​

Now connect a mailbox for the agent to monitor:

  1. Go to Configuration > Email Processing > Monitored Mailboxes
  2. Click Add Mailbox

Configure Mailbox​

  1. Email Address - Enter the shared mailbox address
  2. Display Name - A friendly name for this mailbox
  3. Agent - Select the agent you just created
  4. Auto-Reply - Enable if you want automatic responses

Set Up Subscription​

After adding the mailbox:

  1. Click on the mailbox to view details
  2. Click Manage Subscriptions
  3. Click Create Subscription to start monitoring

Note: Creating a subscription requires Microsoft Graph API permissions. If you see an error, contact your Azure admin.

Step 4: Test Email Processing​

Send a test email to verify everything works:

Send Test Email​

  1. From any email account, send an email to your monitored mailbox
  2. Use a simple subject like "Test inquiry"
  3. Include a question in the body

Check Execution​

  1. Go to Operations > Activity > Agent Executions
  2. Within 1-2 minutes, you should see a new execution
  3. Click on it to see:
    • The incoming email
    • Agent's reasoning
    • Generated response
    • Tool calls made

Verify Response​

If auto-reply is enabled:

  • Check the sender's inbox for the response
  • Review the response quality
  • Note any improvements needed for the agent instructions

Common First-Time Issues​

Mailbox Not Receiving Emails​

Symptoms: No new executions appear after sending test email

Solutions:

  1. Verify the subscription is active (green status)
  2. Check the mailbox permissions in Azure AD
  3. Ensure the email isn't being caught by spam filters

Agent Errors​

Symptoms: Execution shows "Failed" status

Solutions:

  1. Click the execution to see error details
  2. Check LLM provider status and API key
  3. Verify the agent has necessary tools assigned

Poor Response Quality​

Symptoms: Agent responses don't match expectations

Solutions:

  1. Refine the agent instructions
  2. Add more context about your business
  3. Use the AI improvement feature for suggestions

Next Steps​

Now that basic email processing works, consider:

  1. Add VIPs - Set priority levels for important contacts
  2. Create More Agents - Specialize agents for different email types
  3. Configure Search - Add knowledge bases for agents to reference
  4. Set Up Ignore Rules - Filter out unwanted emails