Expand description
What an agent can do — a categorized snapshot for discovery UIs, routing, and debugging.
Every field is optional and omission means not declared, which is not the same as unsupported. Agents fill in only what they mean to advertise.
Structs§
- Agent
Capabilities - A typed, categorized snapshot of what an agent supports.
- Execution
Capabilities - Execution controls and limits.
- Human
InThe Loop Capabilities - Human-in-the-loop support.
- Identity
Capabilities - Agent identity and metadata.
- Multi
Agent Capabilities - Multi-agent coordination.
- Multimodal
Capabilities - Multimodal input and output support, split so clients can query each side independently.
- Multimodal
Input Capabilities - Modalities the agent accepts as input.
- Multimodal
Output Capabilities - Modalities the agent can produce.
- Output
Capabilities - Output formats the agent can produce.
- Reasoning
Capabilities - Visibility into the agent’s reasoning.
- State
Capabilities - State and memory handling.
- SubAgent
Info - A sub-agent a parent agent can invoke.
- Tools
Capabilities - Tool-calling support, and the tools the agent brings itself.
- Transport
Capabilities - Transports the agent speaks.