Skip to content
Rival Docs

What is a Rival Agent

Overview

A Rival Agent is a configurable AI assistant you build inside your organization in the Rival app. Each agent has its own name, icon, persona, and guardrails. A Rival Agent can also call tools, read from your knowledge sources, connect to third-party services like Gmail and Slack, delegate to sub-agents, and run on a schedule via Rituals.

You can find and create Agents from the Build/Studio area or the My Tools page of the app. Each agent has an editor at cortexone.rival.io/rival-agent/{orgSlug}/{id} and a public trial chat through the Marketplace.

What a Rival Agent can do

Every Rival Agent starts from a persona and guardrails and can be given as much capability as the job needs:

  • Tools - call functions to do work, attached from the Marketplace or your own org.
  • Knowledge - read from uploaded files (PDFs, docs, text) when answering.
  • Connectors - act in third-party services via OAuth (Gmail, Slack, Google Drive, and more).
  • Sub-agents - delegate part of a task to another agent.
  • Rituals - run on a schedule, on its own, without anyone in the chat.

You can keep an agent purely conversational - just persona and guardrails - or build it up into something that drives a process end-to-end. It’s the same agent either way; you attach only what you need.

How an agent differs from a tool

A tool is a single function that takes input, runs code, and returns output. An agent wraps a model with a persona and decides which tools to call to satisfy a request. Agents talk to users in natural language; tools are called explicitly with a structured payload.

Agents use tools. Tools never call agents. An agent can also call other agents as sub-agents - one level deep, no recursion.

What you configure on an agent

Every Rival Agent has the same editor layout. The sections are:

  • Name, Description, Instructions & Theme - the very basic identity details.
  • Persona & Guardrails - tone, working style, values, and what’s allowed or off-limits.
  • Tools & Memory - attached tools, uploaded knowledge sources, and OAuth connectors.
  • Sub-agents - other agents this one can delegate to.
  • Rituals - scheduled prompts the agent runs on its own.
  • Connectors - 3rd party tools to help you integrate more platforms with automation.

Publishing is done from the settings option inside the agent. You pick Private, Organizational, or Public visibility, bump the version, and add release notes. Public agents go through marketplace review.

Next steps