What it does
The server provides tools to start a Claude Code session in an absolute working directory, send a prompt and wait for the session to go idle, or send raw input without waiting. It also reports session status, pending permission prompts, todos and transcript lines, and lists sessions it spawned or discovered. A separate claudep_* family drives non-interactive claude -p sessions and returns structured JSON, tool calls, usage and cost.
It suits cases where one agent needs to hand work to other Claude Code sessions and follow their progress, including long tasks that are sent without waiting and checked later. Tool calls are paged, so status polling stays light, and permission prompts are surfaced to the caller to approve or deny rather than bypassed.
Running it requires Node 18 or later and the claude CLI on the PATH. Headless sessions work on any platform. Setting CLAUDE_WRAP_BACKEND=tmux runs sessions in tmux, where they keep running after the server stops, can be attached to from a terminal, and can be adopted by other servers. Session lineage is recorded from TASK_ID and AGENT_ID when they are set, and a registry can be written for a supervisor.
Limitations and requirements
Requires Node.js 18 or later and the claude CLI available on the PATH. Attaching to visible terminal windows is Windows-first, while the tmux backend is for Linux and macOS. Sessions run under tmux persist after the server shuts down unless CLAUDE_WRAP_TMUX_ON_EXIT=kill is set.
Safety
Our team has not reviewed this resource. Only automated rules checked it; resources with high-risk signals are not published. These are signals, not a guarantee that installing it is safe. Try it in a sandbox project first.
- Runs code or changes an assistant's behaviour
Sources
- Official MCP Registry first seen 13 Nov 2026 · last seen 13 Nov 2026 · its page on the source
io.github.Alex-Kaff/claude-wrap-mcp
MCP server that spawns and drives Claude Code sessions via claude-wrap.
v0.1.0As shown on registry.modelcontextprotocol.io when we last read it