Follow Discord
Sweep 22 Oct 2026 · 04:32Z Build v2.1.308 527 read Stable v2.1.308 Latest v2.1.308 Next v2.1.308 Feeds RSS JSON llms.txt llms-full.txt Unofficial
HomeReleasesv2.1.300entry #38

Claude Code v2.1.300 ·

MCP task endings are saved to disk and labelled for SDK programs

The outcome of an MCP server's task is now saved until the model has read it, and its end notification carries taskKind 'mcp-task'

Group of 2 Use it now Improvements
JSON All of v2.1.300
Use it nowTier: how much it should matter to you
3Useful: my rating, 1 to 5
2Signal: worth watching, 1 to 5
SDKArea: what it touches
ImprovementsKind: in v2.1.300,
What probably matters to youSection of the release

What

MCP (Model Context Protocol) is the way outside tools connect to Claude Code, and some MCP servers run long tasks of their own. How their endings are handled changed:

  • A new record on disk stores each MCP task's final outcome: completed, failed or cancelled. It is kept until the conversation transcript is confirmed saved and the model has read the notification, then deleted. Rewinding the conversation to before the call that started the task also deletes it.
  • The schema used by programs built on the Agent SDK gains an optional taskKind field with the value 'mcp-task'. It appears only on a notification that an MCP server's own task has ended, not on every task notification, and the schema marks it @internal.

Why

The result of a long MCP task is kept until the model has actually seen it, so it may survive a session resume. Programs reading these notifications can tell an MCP server's own task ending apart from other task endings.

See this entry in the whole of v2.1.300 →

Feedback