<aside>

Copyl turns AI into governed system users that can act, integrate, and automate business processes inside your existing software stack.

</aside>

Concepts

This chapter introduces the core concepts behind Copyl AI Agents (CAIA). These concepts form the mental model developers need to design, configure, integrate, and operate agents inside the Copyl platform.

Copyl does not treat AI as a feature or chatbot. AI is a first-class system actor.

AI Agents Are Users

In Copyl, an AI Agent is a User.

This is a deliberate architectural choice with several consequences:

• Agents authenticate and exist in the same identity system as humans
• Agents participate in teams
• Agents are assigned roles and permissions (RBAC)
• Agents hold licenses
• Agents generate audit logs
• Agents own tasks and can be task assignees

From the platform’s perspective, there is no special-case execution path for agents. They operate inside the same governance model as the rest of the system.

Agent Types

Agents are flagged as Agent users and can have one or more behavioral classifications, for example:

• Collaboration Agent
• Learning Agent
• Proactive Agent
• Reactive Agent

Agent type affects how and when the agent is allowed to act (for example, trigger-based vs user-initiated), but not its fundamental permissions.

Agent Profiles

An Agent does not have a single personality or configuration.

Instead, it has Agent Profiles.

An Agent Profile defines:

• Instructions / system prompt
• LLM model selection
• Temperature and creativity controls
• Input expectations
• Output schema and guarantees
• Tool access
• Guardrails and constraints

Key properties:

• An agent can have multiple profiles
• Profiles are versioned
• Profiles can target different models
• Profiles can be switched dynamically by the platform or workflow

This allows the same agent to act differently depending on context, task, or risk level.

Rules and Governance

Agents operate under multiple layers of rules:

1. Global Rules – defined at platform level
2. Enterprise Rules – business-specific constraints
3. Agent Rules – agent-specific behavior limits
4. Profile Rules – prompt-level constraints

Precedence is strict:

Global rules always override agent and profile rules.

This ensures that agents can never violate organizational or regulatory boundaries, regardless of prompt configuration.

Read more:

Understanding Prompts and Guardrails in Copyl AI Agents

Triggers and Events

Agents react to events, not polling.

Examples:

• A contract is uploaded
• A task is created or updated
• A record changes state
• An email is received

Triggers create Tasks or Actions that agents can execute based on their permissions and profiles.

This makes agents predictable, observable, and auditable.

<aside> 👉🏼

If you have Triggers defined in the Integration Platform, an Agent can start working directly when the event has happened.

</aside>

Tasks, Calendar and Inbox

Every agent has:

• An inbox (email-like entry point)
• A task list
• A Calendar

Agents:

• Create tasks for humans
• Are assigned tasks by workflows
• Have recurring things todo in the calendar
• Execute tasks autonomously when allowed

This aligns agent activity with human workflows instead of bypassing them.

Tool and API Execution

Agents can execute tools and APIs exactly like any other system actor.

Examples:

• Calling internal Copyl APIs
• Executing integrations via the Integration Platform
• Posting data to external systems (ERP, CRM, Logistics, Payments)

Tool access is explicitly granted per profile and audited.

Reversibility and Learning

Agent actions are reversible.

If an agent performs an incorrect action:

• A user can mark the action as incorrect
• Provide a reason
• The agent rolls back the change
• The feedback is stored as learning context

This is a core safety and trust mechanism.

<aside> 👉🏼

</aside>

Onboarding and Configuration

When an agent is recruited into an enterprise:

• The base agent is cloned
• Enterprise-specific data is collected
• Business rules, policies, and documents are attached
• Profiles are activated or restricted

There are two ways to start using AI Agents in Copyl.

Building Agents for the Marketplace

If you want to build AI Agents from scratch and publish them to the Copyl Marketplace, you need a Developer License. This applies when you define your own agent logic, profiles, tools, tiers, and pricing, and intend to distribute or monetize the agent.

Contact [email protected] if you want to be a Partner and we will onboard you.

Onboarding an existing Agent from the Marketplace

If you only want to use AI Agents, no Developer License is required. In that case, you select an existing agent from the Copyl Marketplace and onboard it into your enterprise by providing configuration, policies, and business context. The agent is then cloned and operates under your rules, permissions, and licenses.