Agents
AI agents that investigate, reason, and act on your behalf
Agents are AI-powered intelligence units that reason and act on your behalf. Use them to investigate alerts, triage incidents, and take action across your connected data, each with its own instructions, tools, and memory.
Agents are available on paid plans.
Adding agents
Install a pre-built ready to use agent or create a custom agent built for your team's specific workflows and needs. Creating, editing, and managing agents requires Owner or Admin access.
To add an agent, go to Agents and click Add agent. You can also create an agent directly from Sola AI by describing the agent you need in the chat.
Installing pre-built agents
Browse ready-made agents from Sola, each designed and tested for a specific security domain. Installing one makes it immediately available to everyone in your workspace.
Available agents
Creating custom agents
Custom agents give you full control over how an agent thinks, what it can access, and how it behaves. Each custom agent has its own instructions, tools, and skills, and is available in your workspace to your entire team.
Tools
Web search - Search the web for up-to-date information
Canvases - Generate and save visual canvases
Alerts - Set up recurring alerts based on findings
Workflows - Create and modify automated workflows
Slack - Send messages and interact with Slack
Jira - Create and update Jira issues
AbuseIPDB - Query IP reputation data from AbuseIPDB
Shodan - Search internet-connected devices via Shodan
Agents always run within the scope of the chat or project they're called from, and can only access the tools you explicitly allow. They do not own data.
Skills
Skills are domain-specific instruction sets that shape how an agent handles specific scenarios. Assigning skills to an agent extends its knowledge beyond its core instructions, without having to rewrite them.
Select from skills already installed in your workspace, or add new ones.
Disabling skills
To disable a skill, remove it from any active agents first.
Instructions
Instructions are the agent's system prompt. They define its role, what it should and shouldn't do, and how it should communicate. The more specific your instructions, the more consistently the agent will behave.
Role & responsibility: The agent's primary goal and area of focus.
Constraints and limitations: What the agent should and should not do.
Response style: How the agent should communicate and format its output.
Invoking an agent
When you @mention an agent in chat, it receives your message, reasons over it using its approved tools and connected data, and returns a response in the thread. Agents run asynchronously. You can continue your conversation while the agent works.
To invoke an agent manually, type @ in the chat input to open the agent picker. Select the agent, add your message, and send.
You can continue your main chat while the agent works. When the agent is ready, its response will appear in the thread referencing your original message.
Active agents When an agent is running in the chat, an indicator bar appears above the chat input, allowing you to monitor active sessions.
To run an agent automatically, use Routines to schedule it, instead of invoking it manually each time. This is useful for recurring investigations or ongoing monitoring.
Agent memory
Agents remember context across chat sessions. Memory is a persistent store the agent reads and writes across conversations, it retains relevant information from past interactions and applies it to new ones, without you having to repeat yourself.
The agent updates its memory independently when it determines that information is worth keeping, such as confirmed findings, preferences, or decisions made during an investigation. You can also explicitly ask an agent to remember something.
Managing agents
The Agents page gives you a workspace-wide view of all installed and custom agents. From here you can see what's active, control which agents are enabled, and manage settings for each one.
To manage your agents, go to Agents.
From the Agents page you can:
See all installed and custom agents at a glance.
Enable or disable any agent.
Try in project, Try in chat, Edit, Duplicate, or Delete.
Add new agents.
Agents, workflows, skills, and routines
Agents think. Workflows orchestrate. Skills guide. Routines automate.
Agents reason dynamically and act on your behalf across connected tools and data. @mention them in chat to investigate, triage, and take action. They are best for open-ended investigations and flexible tasks.
Workflows execute a predefined sequence of steps in a specific order. Trigger them manually or on a schedule to automate repeatable tasks. They are best for repeatable, structured automations.
Skills shape how an agent or Sola AI behaves in a specific domain. Assign them to an agent or apply them in chat to extend its knowledge. They are best for specializing an agent's focus or expertise.
Routines run an agent automatically on a schedule or trigger. Configure one with a task, an agent, and a project scope, and Sola runs it for you. They are best for recurring investigations where the path to the answer may vary.
FAQs
How are agents different from workflows?
Agents and workflows are complementary, not interchangeable. Agents reason dynamically: given a question, they decide how to investigate and respond. Workflows execute a predefined sequence of steps in a specific order. Use agents for open-ended investigations and flexible tasks. Use workflows for repeatable, structured automations where each step is defined in advance.
How are agents different from skills?
Skills are instruction sets that shape how an agent (or Sola AI) behaves in a specific domain. An agent is the actor, with its own role, memory, and approved tools. A skill is a specialization applied on top. You can assign multiple skills to a single agent to broaden what it knows.
Does the agent have access to all my data?
No. Agents do not own or have standing access to data. When invoked, an agent runs within the context of the chat or project it's called from, and can only access the data sources and connectors available in that context. You control which tools the agent can use when you configure it.
Can I invoke multiple agents in the same chat?
Yes. You can @mention different agents within the same chat. Each invocation runs independently. The active agents indicator bar above the chat input shows which agents are currently running.
What happens when an agent creates an artifact?
By default, agents suggest artifacts (queries, canvases, alerts, or workflows) and wait for your confirmation before saving them when invoked in chat. This keeps your workspace clean and ensures you stay in control of what's created. When an agent runs through a Routine on a schedule, it creates artifacts automatically, since there's no one present to confirm.
Can an agent remember something specific about how I work?
Yes. You can tell an agent to remember something directly in the conversation: "Remember that we always exclude test accounts from user audits." The agent will store this in its memory and apply it in future sessions. Memories are shared across all workspace members who invoke the agent.
Who can create and manage agents?
Owners and Admins can create, add, edit, delete, duplicate, manage, and invoke agents. Contributors can invoke agents via @mention.
Can I run an agent automatically, without invoking it myself?
Yes. Routines run an agent automatically on a schedule you define, using the same instructions, tools, and skills the agent already has. Use routines for recurring investigations or monitoring; use @mention for one-off, interactive requests.
Last updated
Was this helpful?