Feature: show sender device on messages for team AI / multiplayer conversations

[simonvanlaak]

Problem

Two of us are already using T3 Code together through T3 Connect, each from our own machine, contributing to the same thread. It works surprisingly well as a shared AI workspace, but we cannot tell which person sent a message.

T3 already distinguishes agent-sent messages. A similarly small label identifying the sending device would make conversations between two people and an agent much easier to follow.

Proposed first step

Show a subtle sender-device label alongside each human message, for example:

• Simon's laptop · 14:32
• Olivia's laptop · 14:33

Reuse the device/client names T3 Code already maintains and displays in Connections → Authorized clients, rather than adding a separate nickname setting or asking users to name their device again. In the current code, the authorized-client display uses the session's AuthClientMetadata.label, with existing OS/browser or session-subject fallbacks. Message attribution should reuse the existing device identity and naming mechanism, carrying the originating client's label with each new message and showing it consistently to everyone viewing the thread.

This needs to work through T3 Connect, including when both clients use the same login. The label must refer to the sending device, rather than always showing the destination environment/server name. If an existing client label is generic (for example, T3 Code Desktop), improve or reuse the existing device-naming path so the two clients can be distinguished; do not introduce a second message-specific naming system. The example names above are illustrative: use whatever device names T3 already knows.

Existing messages could remain unlabeled. Device labels would identify the originating client, without claiming to be verified person identities.

Why this matters

This would be a small first step toward team AI / multiplayer AI: two people can discuss a task with the same agent in the same conversation and see who contributed each instruction or clarification.

The initial request is just visible message attribution. Separate user accounts, permissions, isolated credentials, simultaneous file editing, presence indicators, and a full team architecture can remain separate proposals.

Related work

• Discussion Let the agent know which device the user is typing from #20111 proposes telling the agent which device a message came from, but explicitly leaves the UI unchanged.
• PR feat(server): tell the agent which device sent each message #20112 implemented related sender metadata but was closed without merging pending maintainer agreement. It also notes that hostnames are unavailable behind T3 Connect, which is why attribution should reuse T3's existing client/device identity instead of relying on reverse DNS.
• Issue RFC: OIDC-backed multi-user self-hosted environments #12278 concerns a much broader multi-user authentication and execution model. This request is limited to attribution within an already shared environment.

I understand feature proposals are normally directed to Ideas discussions; recording this specific shared-thread use case here so it can be triaged or converted as appropriate. No implementation PR is attached.

Drafted with Codex in T3 Code.