Main Agent
The top-level model instance in an agent architecture responsible for dialogue management, task coordination, evidence evaluation and final answer generation. It can delegate subtasks such as retrieval to sub-agents.
Core Facts
Timeline (last 90 days)
Antigravity uses a multi-agent architecture in which the main agent can launch specialized background workers (such as a codebase researcher, database debugger, and image generator) through the Invoke sub-agent capability.
The persistent retrieval design allows the main agent to request corrected retrieval results from the same sub-agent.
In the hierarchical agent architecture of main agent and sub-agents, sub-agents are suitable for being handled by small, slow local models, while the main agent, because it needs long context and strong reasoning, is better handed to cloud models.
After removing the standalone planner, the system internalized planning capability into the main agent's reasoning process, reducing handoff costs between components.
Asynchronous sub-agents mean that the main agent can spawn sub-agents to handle time-consuming operations while it continues to interact with the user or handle other matters.
Debugging uses a three-tier role structure: the primary agent is responsible for defining solutions, breaking down tasks, and final acceptance; Luna handles implementation and fixes; and DeepSeek Flash High is responsible for investigation and review via API.
All Facts (6)
Asynchronous sub-agents mean that the main agent can spawn sub-agents to handle time-consuming operations while it continues to interact with the user or handle other matters.
75%UnverifiedAntigravity uses a multi-agent architecture in which the main agent can launch specialized background workers (such as a codebase researcher, database debugger, and image generator) through the Invoke sub-agent capability.
50%UnverifiedThe persistent retrieval design allows the main agent to request corrected retrieval results from the same sub-agent.
50%UnverifiedIn the hierarchical agent architecture of main agent and sub-agents, sub-agents are suitable for being handled by small, slow local models, while the main agent, because it needs long context and strong reasoning, is better handed to cloud models.
50%UnverifiedAfter removing the standalone planner, the system internalized planning capability into the main agent's reasoning process, reducing handoff costs between components.
50%UnverifiedDebugging uses a three-tier role structure: the primary agent is responsible for defining solutions, breaking down tasks, and final acceptance; Luna handles implementation and fixes; and DeepSeek Flash High is responsible for investigation and review via API.
50%