Claude Code 2.1.259 Adds a Deny-Everything Flag for Headless Bots
Claude Code 2.1.259 adds a --permission-prompts none flag that auto-denies anything requiring a prompt on unattended headless hosts, while auto mode keeps making its own decisions.
What it is
Claude Code 2.1.259 is the latest release of Anthropic's terminal coding agent, following close on the heels of 2.1.257's Containment Escape rule.
What it does
It adds --permission-prompts none for headless hosts running unattended, so anything that would normally trigger a permission prompt is denied automatically instead of hanging the session, while auto mode's own decision-making still runs. It also adds a managedMcpServers setting so organizations can push HTTP/SSE MCP servers to every user (entries naming a local command to run are skipped for safety), plus GitLab merge request recognition in the tool summary and footer badge, and --json output for claude plugin commands.
Why it matters
Unattended CI or scheduled runs used to just stall on a permission prompt with no human around to answer it; this flag turns that failure mode into a clean deny instead of a hang, which matters a lot more once auto mode is the default for everyone.
How to use it
Pass --permission-prompts none when launching Claude Code on a headless CI runner or scheduled job; org admins can configure managedMcpServers centrally to roll out approved MCP servers without per-user .mcp.json edits.