Agent capabilities and availability
Agents can combine instructions, knowledge, tools, triggers, memory, flow logic, Skills, and subagents. The exact capabilities available to you depend on your Glean deployment, administrator settings, permissions, enabled connectors, and model configuration.
Capability summary
- Modes: Auto mode supports adaptive execution. Workflow mode supports explicit steps and flow logic.
- Triggers: Agents can run manually and, where enabled, from schedules, messages, or supported system events.
- Tools: Agents can read from connected sources, analyze information, and take approved write actions. Tool access and confirmation requirements are governed by administrators.
- Knowledge: Builders choose the documents, collections, connectors, or other sources an agent can use.
- Skills and subagents: Auto mode agents can reuse available Skills and supported agents as subagents.
- Independent execution: Independent Agents can use an agent-owned profile, supported channel presence, and approved service credentials. This capability is in beta.
- Developer access: Headless building, MCP, A2A, and related developer capabilities have their own setup and availability requirements.
- Models: Available models and model-key or provider requirements depend on the configuration of your organization.
- Outputs: Agents can return responses and, where enabled, create artifacts or update connected systems.
Availability can differ by deployment and can change as features move through limited availability, beta, and general availability. If a capability is missing, ask your administrator to confirm that the feature, connector, model, and permissions are enabled.
For the detailed, canonical view of feature status, deployment, and dependencies, see the Feature and availability matrix.
To choose the right build path, see Choose how to build your agent.