What is a Rival Agent

An agent is an AI worker that takes action toward a goal. You give it instructions, context, and a set of tools, and it carries out real work — not just answering a question, but doing the task.

A Rival Agent is a customizable AI worker created within your organization to complete specific business goals. Beyond standard text conversations, an agent combines a foundation language model with a distinct persona, safety guardrails, knowledge sources, and system tools to carry out real-world tasks.

Each agent receives its own identity, custom branding, and operational scope. An agent can remain purely conversational or expand into an automated, process-handling system. You configure only the capabilities your specific use case requires.

This page is the full reference. For the one-paragraph version, see Agents.

Core capabilities of an agent

Every agent starts with a base persona and security guardrails, then grows through modular capabilities attached by the builder:

• Functions — executes discrete action capabilities from your organization’s collection or the internal marketplace. See What is a function.
• Knowledge retrieval — references uploaded enterprise documents (such as PDFs, policy manuals, or spreadsheets) to ground its responses in verified company facts. See Knowledge base.
• Enterprise connectors — interacts with third-party software platforms using secure authentication. See Connectors overview.
• Sub-agents — delegates specific sub-tasks to other specialized agents within a multi-agent team setup. See Sub-agents and multi-agent systems.
• Rituals — executes automatically on a scheduled, recurring basis without requiring a human to initiate a chat prompt. See Rituals.

How an agent differs from a function

Understanding the relationship between agents and functions is essential for effective system design:

• Functions are discrete actions — a function is a single, deterministic piece of code. It accepts specific inputs, executes an action, and returns an output.
• Agents are decision-makers — an agent pairs a language model with instructions and decides which functions to call, in what sequence, to achieve a complex goal.
• Interaction style — functions require structured, explicit parameter calls, whereas agents communicate in plain language.
• Execution flow — agents invoke functions to accomplish tasks, but functions cannot invoke agents. Additionally, an agent can delegate complex sub-tasks to sub-agents.

Agent editor configuration sections

When building or editing an agent, configuration is organized into six clear panels:

1. Basic identity — set the agent name, icon, public description, primary instructions, and visual theme.
2. Personality and guardrails — define tone of voice, communication style, permitted behaviors, and explicitly prohibited actions. See Personality and guardrails.
3. Functions and memory — attach specific functions, upload knowledge documentation, and link data sources.
4. Sub-agents — designate other workspace agents available for task delegation.
5. Rituals — configure recurring execution schedules and automated triggers.
6. Connectors — link third-party enterprise system integrations.

For the end-to-end build sequence, see Building agents.

Enterprise governance features

While basic agent mechanics remain consistent across workspaces, enterprise deployments add strict governance controls:

• Curated adoption — team members browse and adopt from an admin-approved catalog rather than an unmonitored public marketplace. See Adopting approved agents.
• Publishing workflows — publishing an agent for team or company-wide use routes through administrative review before becoming available. See Publishing an agent.
• Layered guardrails — organization-wide security policies and approved model lists established by IT admins apply on top of individual agent rules. See Guardrails and allowed models.
• Visibility and versioning — agents support Private, Organizational, or Public publishing scopes, complete with version history and release notes.

Capability comparison matrix

Capability	Standard chatbot	Single function	Agent
Primary function	Text generation	Single code execution	Autonomous task completion
System interaction	Isolated conversation	Direct API call	Multi-system connector access
Delegation	None	None	Can route tasks to sub-agents
Scheduled execution	Manual prompt required	Triggered by script	Autonomous runs via rituals
Governance control	Individual user prompts	API key permissions	Admin guardrails, approvals, and logs

Next steps

• Build and deploy your first AI worker → Adopting approved agents
• Explore multi-agent orchestration → Sub-agents and multi-agent systems
• Configure automated execution schedules → Rituals