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.
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.
Unlike working with a language model directly, an Opal Agent packages everything it needs into a single specialized digital worker.
Every agent starts with a prompt that defines its role, how it should behave, and what it is responsible for. This is the foundation of consistent, predictable output.
Each agent is powered by the model best suited to its work. Choose a model directly and set its parameters, or select a pre-configured Model Config for reuse across agents.
Agents draw on organizational Knowledge to ground their work in the right context: your processes, your standards, your information.
Skills give agents reusable, packaged expertise for specific kinds of work, so an agent can be equipped for a particular responsibility without being rebuilt from scratch.
Tools and Connectors (MCP Connections) let agents interact with external systems and services. Agents do not just answer questions. They take action.
Guardrails are attached as input or output guardrails and define how flagged content is handled: redacted, hashed, masked, or blocked. Every agent operates within defined, controllable boundaries.
Write the agent's prompt, select its model, and attach the skills, tools, connectors, and guardrails it needs for its role. An agent configured with the right context and capabilities behaves consistently, every time, without re-briefing.
Once an agent is in a Space, it has access to everything that Space contains: threads, tasks, flows, and organizational knowledge. The Space provides structure, ownership, and access control for the agent's work.
Assign the agent to execute or review Tasks. Assign it to execute or review steps in a Flow. Add it to a Thread to collaborate in real time. In each case, the agent works from the instructions provided for that assignment.
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.
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.
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.
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.
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.
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.
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.
Three real shapes of agent work, from a single scheduled task to a multi-agent flow.
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.
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.
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.
Granular permissions govern who can configure, deploy, and manage agents across the organization.
Input and output guardrails ensure agent behavior stays within safe, compliant limits, with configurable handling for flagged content.
Every agent action is logged. Operations leaders and compliance teams have a complete record of what agents did and when.
Opal supports ZDR configurations to ensure organizational data is not retained by model providers.
Changes to agent configuration are versioned, so you can track how agents have changed over time.
ISO 27001, ISO 42001, and SOC 2 Type II certifications are underway.
Opal Agents are available on every plan. Configure an agent, add it to a Space, and put it to work in minutes.
