What
Claude can send messages to other agents, helpers working on parts of a task, with the SendMessage tool. Three things changed:
- Stopped agents: messages to some agents that were stopped or cleared away now get the reply that the agent "was stopped by the user and was not resumed". Before, only agents stopped directly by the user got this reply.
- Plugins: a plugin can now refuse to resume an agent.
- Queued messages: when a message is queued for a running agent, the reply can add that if the agent finishes first it is resumed with the message and reports back, so Claude should wait for that rather than send again.
Why
Claude is less likely to send the same message twice, and it is told plainly when an agent you stopped will not pick work back up.
Names in the bundleSendMessage
The entry above is what we published on the day. These lines were added later, as Anthropic's own pages caught up, and they sit beside the original rather than replacing it.
Confirmed since
Anthropic's documentation has since written up SendMessage, on Agent SDK reference - TypeScript.
A `peer` origin identifies which agent sent the message: an in-process [teammate](/docs/en/agent-teams) sending to `main` with `SendMessage`, or a [cross-session peer](/docs/en/cross-session-messaging), another of your Claude Code sessions…agent-sdk/typescript see the edit
Two sources agreeTwo things we can check say the same as this entry.
Anthropic's documentation agrees
Anthropic's documentation has since written up SendMessage, on Agent SDK reference - TypeScript.
Anthropic's release notes agree
Fixed SendMessage not finding a subagent by name after the CLI restarted and resumed the session (names given before the last compaction…