About Glean Agents
Glean Agents are reusable AI solutions that combine instructions, knowledge, tools, and context to complete work. An agent can gather information, reason over inputs, make decisions, and take actions across the applications connected to Glean.
Agents can run when a user starts them, on a schedule, or when a supported event occurs. They can return an answer, create a draft, update a connected system, or continue a workflow until it reaches a defined outcome.
Who uses agents?
Agents support three groups:
- Builders create and improve agents for team or department workflows. Builders can include operations teams, IT teams, developers, and subject-matter experts.
- Consumers run agents that teammates or administrators have shared. They use agents to complete repeatable work without designing the workflow themselves.
- Administrators manage access, tools, identities, publishing, and oversight for agents in their organization.
What can agents do?
Use an agent to combine several kinds of work into one outcome. For example, an agent can:
- Research a customer account, summarize recent activity, and prepare a briefing.
- Review a support ticket, find the relevant procedure, and draft next steps.
- Inspect a pull request, compare it with engineering guidance, and generate a review summary.
- Collect meeting decisions, identify action items, and prepare follow-up messages.
- Monitor an approved channel or system event and route work to the right person or tool.
These examples use the same basic model: an agent receives a goal, uses the knowledge and tools available to it, and produces an outcome. The agent's trigger, steps, flow, memory, and subagents determine how that work runs. See How agents work for the mental model.
When should you use an agent?
Agents are a good fit when a task has a clear outcome but requires multiple steps, judgment, connected systems, or recurring execution. For simpler needs, a search experience, a tool, a workflow, or a reusable Skill may be a better fit. See When to use an agent and Agents versus Skills.