Evidence-backed facts
- Agno agents can receive functions, toolkits, MCP tools, and other callable capabilities with runtime hooks and controls. informative
- The agent runtime interprets a model tool call, invokes the selected implementation, and returns the result to the conversation. informative
- Toolkit credentials, cache, retries, confirmation, stop conditions, filesystem and network access remain application-owned controls. informative
- Adding a toolkit does not prove a tool is safe, a call was authorized, an external action completed, or an agent service is deployed. informative