Agents

Agents are the workers behind the scenes. They execute the steps defined by AI Apps and, when needed, collaborate through Agent‑to‑Agent (A2A) requests to share context, perform checks, or return results.

Agent types

Edge Agents (local)

Run directly in the browser with compact, embedded memory. They offer low latency, preserve privacy, and excel at focused, user‑driven tasks.

Example:

• Dana — strategist and tactical executor for quick actions

Cloud Agents (heavy)

Hosted online for higher compute needs. They handle deep analytics, research, and large‑scale retrieval that exceed local resources.

Example:

• Alaric — crypto research and analysis

Coordination and safety

Through A2A, agents can request analysis or validation from each other before an action proceeds. All agents are sandboxed per app and session, with explicit user approvals at critical steps.