Creating & Configuring Agents | FieldCamp

Build AI agents from the marketplace or from scratch. Configure instructions, choose an AI model, connect tools, and test before activating your agent.

There are two ways to create an agent in FieldCamp: pick a pre-built template from the marketplace, or build one from scratch. Either way, you end up with a fully customizable agent you can tweak to match your business.

Create from the Marketplace

The fastest way to get started:

Go to Agents in the left sidebar
Click Marketplace at the top of the page
Browse by category (Communication, Scheduling, Financial, Reporting, Operations, Research)
Click on an agent template to preview what it does
Click Use This Agent
Customize the name, instructions, and connections to fit your business
Click Activate when you are ready

Marketplace agents come pre-configured with sensible defaults. You can use them as-is or change anything before activating.

Create from Scratch

If you need something the marketplace does not offer:

Go to Agents in the left sidebar
Click + New Agent in the top right
Fill in the configuration (covered below)
Test in the Agent Chat panel
Set the status to Active when ready

Agent Configuration

Every agent has these settings:

Name

Give your agent a clear, descriptive name so you and your team know what it does at a glance.

Good names: "Follow-Up Caller," "Invoice Reminder," "After-Hours Receptionist"

Avoid: "Agent 1," "Test," "My Agent"

Instructions

This is the most important part. Instructions tell the agent exactly what to do, how to behave, and what rules to follow. Write them in plain language — the agent understands natural conversation.

A strong set of instructions covers:

What the agent does — "You call clients 3 days after their job is completed to check satisfaction"
How it should behave — "Be friendly and professional. Keep calls under 2 minutes"
What rules to follow — "If the client reports an issue, create a follow-up request in FieldCamp"
What to avoid — "Never discuss pricing or offer discounts"

The more specific your instructions, the better the agent performs. Instead of "handle customer calls," write "answer incoming calls, greet the caller by name, ask how you can help, and log a summary of each call."

AI Model

Choose the brain that powers your agent. Different models have different strengths:

Model	Best For	Speed	Detail
GPT-4o	Complex tasks that need deep reasoning	Moderate	Highest
Claude Sonnet	Detailed, nuanced responses and longer conversations	Moderate	High
GPT-4o Mini	Fast, simple tasks where speed matters more than depth	Fast	Good
Claude Haiku	Quick responses for straightforward tasks	Fastest	Good

Rule of thumb: Use GPT-4o or Claude Sonnet for agents that handle sensitive client interactions (calls, emails). Use Mini or Haiku for internal tasks like status updates and notifications where speed is the priority.

Connections

Connections determine what tools the agent can access. A follow-up caller needs Twilio to make phone calls. An email reminder agent needs Gmail to send messages.

See Agent Connections for the full list and setup instructions.

Status

Draft — the default when you first create an agent. It will not run until you activate it
Active — the agent is live and handling tasks
Paused — the agent is stopped but keeps all its settings

Testing Your Agent

Before activating, use the Agent Chat panel on the right side of the agent configuration page. This lets you have a real conversation with the agent to make sure it behaves the way you expect.

Things to test:

Does it follow your instructions?
Does it respond in the right tone?
Does it handle edge cases (angry client, unusual request)?
Does it use its connected tools correctly?

Make adjustments to the instructions based on what you see, then test again.

Real-World Agent Examples

Follow-Up Caller

What it does: Calls clients 3 days after job completion to check satisfaction.

Setting	Value
Name	Follow-Up Caller
Model	Claude Sonnet
Connections	Twilio, FieldCamp Data
Key instruction	"Call the client, ask if they're happy with the work, and log the response. If they report an issue, create a new request in FieldCamp."

Invoice Reminder

What it does: Sends friendly payment reminders for overdue invoices.

Setting	Value
Name	Invoice Reminder
Model	GPT-4o Mini
Connections	Gmail, FieldCamp Data
Key instruction	"Send an email reminder for invoices that are 7+ days overdue. Be polite but clear about the amount owed and due date. Include a payment link."

Appointment Scheduler

What it does: Handles booking calls and texts from clients.

Setting	Value
Name	Appointment Scheduler
Model	GPT-4o
Connections	Twilio, Google Calendar, FieldCamp Data
Key instruction	"When a client calls or texts to book an appointment, check available slots on the calendar, confirm the date and time with the client, and create the visit in FieldCamp."

Tips for Better Agents

Be specific in instructions — vague instructions lead to unpredictable behavior
Start with one task per agent — an agent that does one thing well is better than one that does five things poorly
Test thoroughly — spend time in the Agent Chat before activating
Review and refine — check your agent's performance after the first week and adjust instructions based on what you see
Use the right model — do not use a powerful model for a simple task, and do not use a lightweight model for complex conversations

Not sure where to start? Browse the Agent Marketplace and activate a pre-built agent. You can always customize it later.

Next Steps

Agent Connections — connect your agent to Twilio, Gmail, Google Calendar, and more
Voice Agents — set up agents that handle phone calls
Agents in Workflows — use agents as steps in your automations
AI Skills — see everything the AI can do in FieldCamp

Creating & Configuring Agents | FieldCamp

On this page