AI/TLDR

Anthropic · 2026-09-02 · major

Claude Code 2.1.259 — admins can push MCP servers to every user

Claude Code 2.1.259 adds a managedMcpServers setting so an organization can hand HTTP and SSE MCP servers to all of its users at once. A new --permission-prompts none flag makes unattended headless hosts deny prompts instead of hanging.

GitHub card for the anthropics/claude-code repository
GitHub

The 2.1.259 release moves MCP configuration up to the administrator, and gives headless runs a way to refuse prompts instead of hanging.

Quick facts

Version2.1.259
ReleasedSeptember 2, 2026
New managed settingmanagedMcpServers (HTTP/SSE servers for all users)
New flag--permission-prompts none
Scope changeallowedMcpServers now governs only user-added servers
GitLabRecognizes glab mr create/merge/close/reopen/note/update
Headless startUp to 50ms faster

What is it?

A new managedMcpServers managed setting is the headline of Claude Code 2.1.259: organizations can supply HTTP and SSE MCP servers to every user from one place. The same release adds --permission-prompts none for unattended headless hosts and a --json flag on claude plugin validate for machine-readable validation reports.

How does it work?

MCP servers in Claude Code previously lived in local, project and user scopes that each developer maintained. managedMcpServers adds an enterprise scope on top, delivered through managed configuration. To keep the two apart, allowedMcpServers was narrowed so it now filters only servers a user added themselves. The headless path also got faster, with session start improved by up to 50ms.

Why does it matter?

Rolling an internal tool out to a whole engineering org stops being a documentation exercise. An admin publishes the server once and every Claude Code user has it, which also means the tool list can be audited centrally. On the automation side, a headless job that hits an unexpected permission prompt used to sit there; now it can be told to deny and move on.

Who is it for?

platform teams and anyone running Claude Code unattended

Frequently asked questions

What problem does managedMcpServers solve for an organization?
Before 2.1.259, every Claude Code user wired up their own MCP servers. The managedMcpServers managed setting lets administrators provide HTTP and SSE MCP servers to all users centrally, so a company connector reaches every developer without each person editing config. It sits alongside Claude Code's existing local, project and user scopes.
When would I use --permission-prompts none?
Claude Code's --permission-prompts none flag is meant for unattended headless hosts. Instead of waiting on a human who is not there, it automatically denies permission prompts, so a CI job or a scheduled run fails cleanly rather than stalling. Interactive sessions keep their normal prompt behavior.
Did the behavior of allowedMcpServers change?
Yes. In Claude Code 2.1.259 allowedMcpServers now governs only user-added servers. That separates the allow list a person controls from the servers an administrator distributes through managedMcpServers, so an org-provided connector is not filtered out by an individual's own list.
What did this release fix for teams running several sessions at once?
Claude Code 2.1.259 fixes concurrent sessions silently reverting each other's changes to ~/.claude.json, which previously lost configuration edits without warning. The release also repairs managed settings going unenforced because of parsing errors, and stops managed forceRemoteSettingsRefresh being ignored at startup.

Try it

npm install -g @anthropic-ai/claude-code@2.1.259

Sources · 2 outlets

Tags

  • claude-code
  • anthropic
  • mcp
  • coding-agents
  • developer-tools
  • enterprise
  • headless
  • gitlab
  • vscode

← All releases · Learn AI