Feature and availability matrix
This matrix is the canonical detailed reference for agent feature availability. Availability can vary by deployment, user role, enabled feature, connector, model provider, and rollout stage. Confirm the current options in your Glean environment before you design or publish an agent.
| Capability | Availability and execution model | Key dependencies |
|---|---|---|
| Auto mode | Default starting mode for adaptive, outcome-oriented execution. | Agentic model, enabled Auto mode access, and available tools and knowledge sources. |
| Workflow mode | Explicit, step-by-step execution with branches, loops, and hand-offs. | Agent Builder access and the tools, triggers, and models used by the workflow. |
| Agent Builder | Visual entry point for creating, refining, testing, and publishing agents. | Permission to create or edit agents; available features depend on administrator settings. |
| Natural-language Workflow creation | Generates initial Workflow mode steps from a structured description. | Agent Builder access and an enabled agentic model. |
| Templates | Starting assets for common workflows. | Templates, connectors, tools, and permissions available in the deployment. |
| Independent Agent | Beta execution model with an agent-owned profile and operation without an active user session. | Beta access, supported triggers or channels, administrator approval, and possibly service credentials. |
| Triggers | Manual execution is available where the agent is accessible. Scheduled, message, and system-event triggers depend on deployment support. | Trigger feature access, event source, filters, activation, and administrative settings. |
| Headless builder | Code-based path for building and testing Auto mode agents. | Glean plugin, authentication, local agent files, and Auto mode access. |
| MCP | Lets supported external hosts use Glean capabilities through configured MCP access. | MCP support, host configuration, authentication, and administrator controls. |
| A2A | Lets supported agents expose or consume agent capabilities through the Agent2Agent protocol. | A2A support, endpoint configuration, authentication, and administrator controls. |
| Models and model keys | Available models and providers vary by organization and deployment. | Model Hub configuration, provider setup, model access, and any required model keys or credentials. |
| Skills and subagents | Available Auto mode agents can reuse Skills and supported agents as subagents. | Skills rollout, access to the Skill or agent, and Auto mode support. |
| Artifacts and write actions | Output formats and write actions depend on the enabled capabilities and connected systems. | Artifact feature access, connector tools, permissions, and confirmation or approval settings. |
For setup details, see Configure LLMs, MCP for agents, Expose agents over A2A, and Agent identity.