Zsh Tool
Give Claude Code a zsh MCP shell with PTY password entry, hang-resistant execution, and A.L.A.N. session memory for real terminal workflows.
Overview
zsh-tool is a MCP server for the Build phase that exposes PTY-capable zsh to Claude Code with NEVERHANG execution and A.L.A.N. session memory.
What is this MCP server?
- Zsh-backed MCP shell tailored for Claude Code terminal workflows
- PTY mode so the agent can handle interactive prompts including password entry
- NEVERHANG behavior aimed at avoiding stuck long-running shell sessions
- A.L.A.N. SQLite-backed context via optional ALAN_DB_PATH environment variable
- PyPI zsh-tool-mcp v0.2.1 with uvx runtime hint and stdio transport
- Server version 0.2.1
- PyPI identifier zsh-tool-mcp with uvx runtime hint
- Stdio MCP transport
Community signal: 4 GitHub stars.
What problem does it solve?
Standard agent shell tools hang on interactive prompts and cannot type passwords, so real deploy and dev commands fail mid-workflow.
Who is it for?
Indie developers who live in Claude Code and need dependable, interactive terminal access for installs, scripts, and ops commands.
Skip if: Builders who only need read-only file edits or who forbid agents from running shells on security grounds.
What do I get? / Deliverables
Your agent runs zsh through MCP with PTY interactivity and more reliable long-running command behavior across the solo-builder journey.
- PTY-capable zsh command execution from the agent
- More reliable long-running shell sessions with NEVERHANG-oriented behavior
- Optional persistent A.L.A.N. command context across MCP turns
Recommended MCP Servers
Journey fit
Agent tooling is the canonical shelf because the server exists to extend how coding agents run shell commands during implementation. Agent-tooling subphase covers MCP bridges that replace brittle raw exec with interactive zsh suited to Claude Code sessions.
How it compares
Interactive zsh MCP bridge, not a hosted CI runner or a static lint skill.
Common Questions / FAQ
Who is zsh-tool for?
zsh-tool is for solo builders using Claude Code who want MCP-driven zsh with PTY support instead of fragile non-interactive shell snippets.
When should I use zsh-tool?
Use it whenever your agent must run local zsh commands, pass interactive prompts, or reuse terminal context via A.L.A.N. during build, ship, or operate work.
How do I add zsh-tool to my agent?
Add the PyPI package zsh-tool-mcp as a stdio MCP server (uvx runtime hint), set ALAN_DB_PATH if you want A.L.A.N. persistence, then call its shell tools from Claude Code.