Unverified50% confidenceFactExact time
Functions declared by the agent in the Catalog (such as opening a URL or validating input) are not executed by the agent itself; they are sent back to the client as action messages, and the client decides how to handle them, leaving the authority to execute side effects on the application side.
1
Sources
50%
Confidence
Long-term
Relevance
10/30/2026
First Seen
Sources
Generative UI Comes to Android: A2UI Lets Agents Build Native Interfaces in Real Time
youtubeAndroid Developers10/30/2026
Related Entities
Related Claims
UnverifiedWhat an Agent needs is semantic action descriptions (such as "query order status" and "initiate a return"), not a list of low-level functions or endpoints.77% similarVerifiedAn Agent's Action capability relies on standardized interfaces such as Function Calling or Tool Use to convert the model's text output into executable API calls.76% similarUnverifiedAgents access project data and perform actions on behalf of users, so the permission model must be incorporated at the early stage of Agent design rather than added afterward.75% similarUnverifiedon_agent_action fires when the Agent decides to perform an action but has not yet actually executed the tool, so action.log is the Agent's complete thinking before this step, not a post-hoc summary.75% similarUnverifiedThe Agent runs with the permissions of its process, not with the permissions of the user who initiated the prompt.74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1118264API
curl https://kongchang.com/api/v1/knowledge/claims/1118264MCP
get_claim(id=1118264)