// November 13, 2026v1.52.3
Platform
Identity
Troubleshooting your own agent used to need a project-wide
chat:read or log:read grant, the same permission that opens every other member's conversations. Now you can read the raw logs, chats, and risk findings of your own sessions and of the agents you own without either one. More on this in Agent Sessions.
Features
- Read your own and your agents' tool logs without
log:read#8598 - The raw tool-call log and trace endpoints no longer refuse a member who lacks project-widelog:read. They now narrow to the caller's own sessions and the sessions of the agents they own, so you can troubleshoot an agent you're responsible for without a grant that reveals every member's tool arguments and results. (Author: @dennnis-ez) - Platform MCP chat and log tools follow the same ownership rule #8617 -
list_chats,search_chats,read_chat,search_tool_calls, andlist_recent_tool_callsnow cover your own chats and tool calls plus those of agents you currently own, matching the visibility the dashboard already gives. Achat:readorlog:readgrant on the project still widens the view to the whole project. (Author: @dennnis-ez) - Unmask a risk finding on your own sessions without
chat:read#8618 - You can now reveal the matched value of a risk finding on a session you own, or on a session of an agent you own, without holding the organization-widechat:readpermission. The reveal is audited the same way it already was. (Author: @dennnis-ez)
Bug fixes
- Session lists no longer leak metadata past
chat:read#8616 - Session lists used to show per-session details, like a user's email, title, model, and per-chat metrics, for projects you didn't holdchat:readon. They now come only from projects you havechat:readon, and a search without that grant narrows to your own sessions and those of agents you own instead of listing everyone's. (Author: @dennnis-ez)
