Configuring Agents

Agents are highly configurable. This guide covers all configuration areas and how to use them effectively.

Configuration Areas

Agents can be configured in 10 main areas:

  1. Basics — Name, status, avatar
  2. Persona & Tone — How the agent communicates
  3. Working Patterns — When and how the agent works
  4. Human-in-the-Loop — When to involve you
  5. Context & Knowledge — What the agent knows
  6. Specialists — Intent-aware helpers
  7. Policies & Guardrails — Rules to follow
  8. Tools & Integrations — Connected services
  9. Model & Execution — AI model settings
  10. Ops & Telemetry — Logging and monitoring

Basics

Location: Agent Config → Basics tab

Name

  • Display name for the agent
  • Example: "Jordan", "Alex", "Inbox Manager"
  • How you'll refer to the agent in the UI

Status

  • Active: Agent is running and processing work
  • Paused: Agent is temporarily disabled
  • Change status to control when agents work

Avatar

  • Emoji: Quick visual identifier
  • Image: Upload a custom avatar
  • Helps identify agents at a glance

Tags

  • Labels for organizing agents
  • Example: "ops", "email", "billing", "support"
  • Use tags to filter and group agents

Persona & Tone

Location: Agent Config → Persona tab

Tone

Choose how the agent communicates:

  • Professional: Formal, business-like
  • Friendly: Warm, approachable
  • Formal: Very formal, traditional
  • Casual: Relaxed, informal

Traits

Select personality traits:

  • Concise: Brief and to the point
  • Empathetic: Understanding and caring
  • Decisive: Makes clear decisions
  • Customer-focused: Prioritizes customer needs
  • Detail-oriented: Pays attention to specifics

Bio

  • Short description of the agent
  • Used in introductions and first contact
  • Example: "I'm Jordan, your inbox manager. I help organize and respond to your emails."

Signature

  • Email signature template
  • Supports variables: {{agent.name}}, {{identity.handle}}
  • Example: 
    Best regards,
{{agent.name}}
{{identity.handle}}@solomail.io

Style Preferences

  • Max sentence length: Keep responses concise
  • Use bullets: Prefer bullet points
  • Format: Plain text or Markdown

Working Patterns

Location: Agent Config → Working Patterns tab

Timezone

  • Agent's timezone
  • Used for scheduling and time-based decisions
  • Example: "America/New_York"

Working Hours

  • When the agent is active
  • Define hours per day of week
  • Example: Monday-Friday, 9 AM - 5 PM

SLAs (Service Level Agreements)

  • First Response: Target time for first response
  • Follow-up: Target time for follow-ups
  • Example: 30 minutes for first response, 24 hours for follow-up

Off-Hours Behavior

  • Queue: Hold messages until working hours
  • Auto-acknowledge: Send acknowledgment, handle later
  • Delegate: Pass to another agent

Human-in-the-Loop (HITL)

Location: Agent Config → HITL tab

Confidence Threshold

  • Percentage (0-100%) when agent should escalate
  • Low (50%): More escalations, more control
  • Medium (70%): Balanced (recommended)
  • High (90%): More autonomy

Intent Triggers

Always escalate for these topics:

  • Legal: Legal matters, contracts, disputes
  • Billing: Payment issues, refunds, disputes
  • Sensitive: Topics you define
  • Custom: Add your own triggers

Approval Mode

  • Hold for Approval: Review before action
  • Send Unless Declined: Act unless you decline

Escalation Routes

Set up escalation paths:

  • Level 1: First escalation point (role, channel, timeout)
  • Level 2: Second escalation point
  • Level 3: Final escalation point

See Human-in-the-Loop for complete details.

Context & Knowledge

Location: Agent Config → Context tab

Datasets

  • Knowledge bases the agent can reference
  • Include/Exclude: Toggle datasets on/off
  • Refresh: Manual, daily, or hourly updates

Retrieval Settings

  • Top-K: How many results to retrieve
  • Min Score: Minimum relevance score
  • Recency Bias: Prefer recent information
  • Citations: How to cite sources

Namespace

  • Context scope: Identity and agent-specific
  • Display-only: Shows context boundaries
  • Example: {identityId}:{agentId}

Specialists

Location: Agent Config → Specialists tab

Allowed Specialists

  • Select specialists this agent can invoke
  • Specialists are defined in Settings → Specialists
  • Each specialist handles specific intents

Priority Order

  • Order specialists by priority
  • Agent tries first specialist, then next if unavailable
  • Example: Scheduling Specialist → Billing Specialist

Intent Mapping

  • Map intents to specialists
  • Example: "scheduling" → Scheduling Specialist
  • Use regex or tags for matching

Auto-Selection

  • Let system choose specialist based on detected intent
  • Agent analyzes intent and selects appropriate specialist
  • Fallback to manual selection if needed

Failover

  • Try next specialist if first is unavailable
  • Ensures requests are handled even if specialist is down
  • Follows priority order

See Specialists for details.

Policies & Guardrails

Location: Agent Config → Policies tab

Policy Packs

  • Attach policy packs defined in Settings
  • Policy packs contain rules agents must follow
  • Examples: Brand style, safety rules, spending caps

Enforcement Level

  • Block: Prevent action if policy violated
  • Require Approval: Escalate for review
  • Warn: Log violation but allow action

Precedence

  • Identity overrides: Identity policies override agent
  • Most restrictive wins: Apply strictest policy
  • Choose based on your needs

Effective Policy

  • View resolved policy: See what rules actually apply
  • Combines identity and agent policies
  • Shows enforcement level

See Policy Packs for details.

Tools & Integrations

Location: Agent Config → Tools tab

Email

  • Provider: Mailgun (default)
  • Folders: Which folders to monitor
  • Threading: How to handle email threads

Calendar

  • Providers: Google Calendar, Outlook, etc.
  • Buffers: Minimum notice, buffer time
  • Templates: Meeting templates

Storage

  • Providers: Supabase, S3, etc.
  • Allowed file types: Which files can be stored
  • Size limits: Maximum file sizes

HTTP/Web

  • Allowed hosts: Which domains can be accessed
  • Methods: GET, POST, PUT, DELETE
  • Headers: Custom headers for requests

Headers & Secrets

  • Vault references: Secure storage for API keys
  • Environment variables: Access to env vars
  • Credentials: OAuth tokens, API keys

Model & Execution

Location: Agent Config → Model tab

Provider

  • Solo: Solo's optimized models (recommended)
  • OpenAI: GPT-4, GPT-3.5, etc.
  • Anthropic: Claude models
  • Custom: Your own model endpoint

Model

  • Specific model to use
  • Example: "gpt-4", "claude-3-opus"
  • Depends on provider

Parameters

  • Temperature: Creativity (0-1, lower = more focused)
  • Max Tokens: Maximum response length
  • Top-P: Nucleus sampling
  • Frequency Penalty: Reduce repetition

Tool-Calling Policy

  • Strict: Only use defined tools
  • Balanced: Use tools when helpful (recommended)
  • Exploratory: Try tools more freely

Budget

  • Daily USD: Maximum spending per day
  • Monthly USD: Maximum spending per month
  • Warn At: Percentage to warn at (e.g., 80%)

Reasoning Visibility

  • None: No explanation
  • Brief: Human-safe rationale
  • Full: Complete reasoning (may be technical)

Ops & Telemetry

Location: Agent Config → Ops tab

Logging

  • Minimal: Errors only
  • Normal: Errors and important events (recommended)
  • Verbose: All events and decisions

Alerts

Configure alerts for:

  • Error Spike: Too many errors
  • Budget Near Limit: Approaching spending limit
  • Backlog: Too many pending items
  • Custom: Your own alert conditions

Shadow Mode

  • Observe Only: Agent observes but doesn't act
  • Draft Only: Creates drafts but doesn't send
  • Useful for: Testing new configurations

Export

  • Config Snapshot: Export agent configuration
  • Useful for: Backup, sharing, version control
  • Premium Feature: Available on paid plans

Saving Configuration

Save Changes

  • Click "Save" button (top right or mobile)
  • Changes are saved immediately
  • Agent updates with new configuration

Discard Changes

  • Click "Discard" to revert
  • Returns to last saved state
  • Useful if you make mistakes

Auto-Save

  • Some changes auto-save
  • Look for save indicator
  • Manual save always available

Configuration Best Practices

Start with Defaults

  • Use default settings initially
  • Let agents learn from interactions
  • Adjust based on results

Configure Gradually

  • Don't change everything at once
  • Test changes before making more
  • Keep what works, adjust what doesn't

Document Changes

  • Note what you changed and why
  • Helps if you need to recreate setup
  • Useful for team members

Test Thoroughly

  • Use shadow mode for testing
  • Review agent actions after changes
  • Adjust based on results

Next Steps

Version: 1.0
Last Updated: November 2025