Agents

Deploy digital workers that get things done.

Opal Agents are purpose-built digital workers that participate in conversations, execute tasks, run workflow steps, and take action in connected systems. Configure once. Put to work everywhere in your organization.

Start for free Request a demo

Agents · networklive

Multi-agent graph · live4 agents · 1 human

orchestrator

router

research

agent

writer

agent

sender

agent

mira

review

agenthuman-in-the-loop

routing · 3 active

The limits of general-purpose AI

A chatbot answers questions. An agent does the work.

Most AI tools require you to re-brief them every time. They have no memory of your organization, no defined role, no connection to your systems, and no way to participate in actual work. They are useful for one-off tasks. They are not built for sustained, reliable execution at scale.

Opal Agents are different. Each agent is configured with a defined role, grounded in organizational knowledge, equipped with the right tools and connections, and governed by guardrails that keep its inputs and outputs within safe, compliant boundaries. It does not need to be re-briefed. It knows its job.

Why Opal Agents

Purpose-built digital workers that scale with your team.

Extends team capacity

Delegate repeatable and specialized work to a digital worker that operates alongside your team. Free your people to focus on higher-value work without adding headcount.

Consistent by design

An agent packages its prompt, model, knowledge, skills, tools, and guardrails into a single specialized worker. It applies the right context and behaves the same way every time.

Stronger through collaboration

Multiple agents can work together on the same effort, each contributing its own expertise. Use an executor-reviewer approach to add a layer of quality checking at every step.

Grounded in your organization

Agents draw on your organizational Knowledge and connected systems. Their work reflects your information, your standards, and your data, not a generic model's best guess.

Governed and accountable

Guardrails, defined permissions, and audit logs keep agent behavior within safe, compliant boundaries. Scale digital work while staying in full control of how it is done.

Built-in quality

One agent executes. Another reviews.

Opal's executor-reviewer approach lets you assign one agent to carry out a task or workflow step and a second agent to review the output before it is accepted. Each works from its own instructions: execution instructions for the agent doing the work, review instructions for the agent checking it.

The result is a layer of quality checking built into the process itself. Issues are caught before output is accepted, without requiring a person to review every result.

This approach is available in Tasks and in every step of a Flow, which makes it easy to apply consistently across multi-step processes.

Task · quality looprunning

executor agent

execution instructions

reviewer agent

review instructions

accuracy oktone okpolicy ok

output acceptedno human review

Better together

Specialized agents, working as one.

Because each Opal Agent can be specialized for a distinct role, several agents can work together on the same effort, each contributing its own expertise, rather than relying on one general-purpose assistant to do everything.

A research agent gathers information. A writing agent drafts the content. A compliance agent reviews the output. Each plays a defined role. Each applies its own knowledge, skills, and guardrails. The result is work that reflects real organizational depth, not a single model's approximation.

Multi-agent collaboration is available in Threads, Tasks, and Flows.

One general assistant

general assistant

researchsame generic context

draftingsame generic context

compliancesame generic context

deliverysame generic context

one prompt for everythingno depth

Specialized agents

one shared effort

researchgathers sources

writingdrafts the work

compliancereviews output

deliveryships the result

own knowledge, skills, guardrails4 specialists

Agents in action

How organizations put agents to work.

Three real shapes of agent work, from a single scheduled task to a multi-agent flow.

Case 01

Customer Success

Renewal preparation

A customer success team was spending hours each week compiling account health summaries before renewal conversations. They created a Customer Success Agent, connected it to their CRM, and attached the team's renewal criteria as Knowledge. A scheduled Task runs every Monday: the agent pulls the latest account data, applies the health criteria, and produces a summary for each upcoming renewal. A Review Agent checks each summary for completeness before it is accepted. What once took hours now runs automatically, and the team spends their time on the conversations, not the prep.

Agents: Customer Success Agent
Renewal Review Agent
Surface: Scheduled Task
Outcome: Hours of prep removed each week

Case 02

Operations

Project status reporting

A program management team needed consistent weekly status reports across a portfolio of projects. They configured a Reporting Agent with their status criteria and connected it to their project data. Each week, the agent runs through a Flow: pulling the latest data, generating the status report, and routing it to a reviewer before it is distributed. The team gets consistent, reliable reporting without the manual overhead.

Agents: Status Reporting Agent
Report Review Agent
Surface: Weekly Flow
Outcome: Consistent portfolio reporting

Case 03

Content

Research, draft, review

A content team uses three specialized agents working together in a Flow. A Research Agent gathers source material and summarizes key points. A Writing Agent drafts the content using the team's standards and style. An Editorial Review Agent checks the draft against brand guidelines before it moves to human review. Each agent is specialized. Each plays a defined role. The team gets further through the production process before a human needs to engage.

Agents: Research Agent
Writing Agent
Editorial Review Agent
Surface: Multi-agent Flow
Outcome: Drafts arrive review-ready

Governed from the ground up

Every agent operates within defined, auditable boundaries.

Role-Based Access Control

Granular permissions govern who can configure, deploy, and manage agents across the organization.

Guardrails

Input and output guardrails ensure agent behavior stays within safe, compliant limits, with configurable handling for flagged content.

Audit Log

Every agent action is logged. Operations leaders and compliance teams have a complete record of what agents did and when.

Zero Data Retention

Opal supports ZDR configurations to ensure organizational data is not retained by model providers.

Versioning

Changes to agent configuration are versioned, so you can track how agents have changed over time.

Certifications

ISO 27001, ISO 42001, and SOC 2 Type II certifications are underway.

Agents are one part of a connected platform

Everything agents need built in.

01SpacesAgents are added into Spaces, which provide organizational structure, ownership, and access control.02FlowsAgents execute and review steps within multi-step workflows.03TasksAgents are assigned as executors or reviewers of individual tasks.04ThreadsAgents participate in real-time conversations alongside people.05KnowledgeAgents draw on Knowledge to ground their work in organizational context.06SkillsSkills equip agents with reusable, packaged expertise for specific responsibilities.07ToolsTools let agents interact with external systems and services to take action.08ConnectorsMCP Connections give agents access to external data sources and platforms.09GuardrailsGuardrails govern agent inputs and outputs and define how flagged content is handled.10ModelsEvery agent is powered by a model; Model Configs package model choices for reuse.

Deploy your first agent today.

Start building with agents.

Opal Agents are available on every plan. Configure an agent, add it to a Space, and put it to work in minutes.

Start for free Request a demo