Claude Code 2.1.318 tightens MCP deny rules and subagent behaviour

Claude Code 2.1.318 makes deniedMcpServers URL entries block Anthropic-provided servers and connectors, and changes how subagents handle flagged messages and long shell commands.

what changed

Claude Code 2.1.318 was listed on 4 November 2026. These are the changes that matter most for teams:

  • deniedMcpServers URL entries now also block Anthropic-provided servers whose upstream URL matches. They also block connectors in cloud sessions and on self-hosted runners, where the session is given the connector’s own URL. To deny a connector otherwise, deny it by name.
  • Subagents now stop, instead of switching models, when “Switch models when a message is flagged” is set to “Ask each time”.
  • Background subagents move a shell command to the background after 3 minutes on all providers, so the prompt cache does not expire while it runs.
  • Claude apps gateway mantle upstreams gain assume_role, for a role in another AWS account and for per-developer AWS cost attribution.
  • /claude-api managed-agents-onboard now gives a new agent’s built-in tools the auto permission policy (“Judge risky actions”) by default.

why it matters

The MCP change affects anyone who uses a deny list as a control. A URL rule that used to miss a connector may now catch it. That is the intended result, but a rule you wrote months ago may now block a tool a team relies on.

The subagent changes affect long agent runs. A subagent that stops where it used to switch models can leave a task half done, and that shows up as a failed run, not a warning.

Platform teams that run the gateway on Amazon Bedrock can now split cost by developer across AWS accounts.

what we would do

  • Review your managed deniedMcpServers entries. Test that each one blocks what you meant, and nothing else.
  • If you set “Ask each time” for flagged messages, check that unattended subagent runs have a plan for stopping.
  • Look at any agent that runs shell commands for longer than 3 minutes, and confirm it copes with the command moving to the background.
  • If you create agents with the onboarding command, read the new default permission policy before you accept it.

sources

Drafted with Claude from the linked sources, reviewed by us.