Zero CLI
A Go terminal coding agent that works with any model and asks permission before commands, network access and writes
- 1,741 ★ on the repository
- Last commit
Broad access
Why this level
- The agent writes files and runs shell commands, and autonomous modes without prompts are a separate choice
- Keeps provider keys and subscription logins in user settings and sends project content to the chosen model
- Plugins and hooks run external commands on beforeTool, afterTool and session events
Install
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/zeroDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The agent writes files and runs shell commands, and autonomous modes without prompts are a separate choice
- Keeps provider keys and subscription logins in user settings and sends project content to the chosen model
- Plugins and hooks run external commands on beforeTool, afterTool and session events
This is third-party code. Review the repository files before installing.
What it does
Zero is a coding agent that runs in your local terminal. It reads the repository, edits files, runs commands and keeps sessions on disk that you can search, resume and roll back with `/resume` and `/rewind`. You pick the model and provider: more than a dozen cloud APIs from OpenAI and Anthropic to DeepSeek and Qwen, local Ollama and LM Studio, or any OpenAI- or Anthropic-compatible endpoint. The zero command opens an interactive interface with slash commands. `/plan` and `/spec` draft a plan before editing, `/loop` repeats a prompt on a schedule and Shift+Tab cycles the permission mode. `zero exec "task"` runs headless and suits scripts and CI. Zero is extended with MCP servers, skills, plugins with hooks and specialist subagents. Project instructions come from AGENTS.md, ZERO.md or .zero/AGENTS.md.
Who it is for. For developers who want a terminal agent like Claude Code with any model of their choice, local sessions and managed permissions.
Good fit when
- You want a terminal agent that works with Ollama, LM Studio or a paid API of your choice and is not tied to one vendor
- You want to run an agent in CI or a script and parse its output in code
- You want control: writes only inside the workspace, a prompt for every command and network call, and a policy check with `zero sandbox policy`
- Your project uses AGENTS.md, MCP servers and skills and you want to wire them into one agent
Not a fit when
- You want an agent inside an editor with a graphical interface: Zero works in the terminal and headless
- A Linux environment without Bubblewrap: the built-in sandbox there relies on it
- You reach a model only through an account subscription: check whether your login method is listed in docs/oauth-subscriptions.md
- You only need a one-off model query: calling the API directly is simpler than setting up an agent
Example request
Run zero exec with an isolated worktree: fix the failing test in ./pkg, run that package tests and show the changesLimitations
- A provider key or a local model is required. Keys are set through environment variables such as OPENAI_API_KEY or the `zero setup` wizard. OpenAI and Anthropic APIs are unavailable from Russia, so check your provider before installing.
- On Linux the built-in sandbox needs Bubblewrap. When no separate helper exists, the main binary takes over its role. macOS needs no extra binary.
- Building from source needs Go 1.26.10 or newer.
- When HTTP_PROXY or HTTPS_PROXY is set, web_fetch sends the target hostname to the proxy, and the address check then depends on the proxy.
- The npm package moved from @gitlawb/zero to @twigpine/zero, and the old name gets no more releases.
- The license is MIT.
How to disable. Remove the package with `npm uninstall -g @twigpine/zero`. Settings, sessions and logins live outside the package in the user config directory; delete them separately. Disable plugins, MCP servers and skills with `zero plugins`, `zero mcp` and `zero skills`.
FAQ
How do I install Zero and where do I start?
Run `npm install -g @twigpine/zero`, then zero. The wizard asks you to pick a provider and model. The script for Linux and macOS is `curl -fsSL https://raw.githubusercontent.com/Twigpine/zero/main/scripts/install.sh | bash`, and Windows has install.ps1. `zero doctor` helps check the setup.
How do I run Zero in CI or a script?
With `zero exec "task"`. For machine processing use `--output-format stream-json`, for a safe attempt `--worktree`, to continue an earlier session `--resume` and for a branch `--fork`. The stream format is described in docs/STREAM_JSON_PROTOCOL.md.
What permissions does Zero have by default?
Reading the workspace is allowed and writes are limited to it. Shell commands, network, destructive and elevated actions require permission. The `--add-dir` flag grants an extra write directory, and autonomous modes are explicit opt-ins.
How do I connect MCP servers and skills to Zero?
Through `zero mcp` and `zero skills`, or through a plugin: a plugin.json manifest in ~/.config/zero/plugins/<name>/ or .zero/plugins/<name>/ can declare tools, hooks, prompts and skills.
Does Zero send my sessions anywhere?
Sessions are stored on disk, searched and resumed locally, and Zero does not upload them as telemetry. Model requests go to the provider you chose.
Discussion
No questions or reviews yet.
Related
CLI toolsOpen-source personal AI assistant on your own machine: answers in Telegram, Slack, Discord and WhatsApp, extended with skills and plugins
An engineering system for coding agents: 68 agents, 292 skills, hooks, memory, rules and the AgentShield security scanner
DeepSeek Harness is an open-source AI agent from DeepSeek where every feature is a plugin. It runs in the browser, on macOS and on Windows.
A self-improving agent from Nous Research with a TUI, messaging gateway, cron jobs and skills it writes itself
An open source coding agent for the terminal and desktop with build and plan modes
Ollama runs open models on your own machine and serves them to agents through a local API compatible with OpenAI and Anthropic.
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Repository stars | Main difference |
|---|---|---|---|
| Zero (this entry) | no | 1,741 | |
| OpenCode | no | 220,388 | another terminal and desktop agent with build and plan modes, while Zero focuses on the sandbox, exec mode and streaming JSON |
| Qwen Code | no | 28,667 | the Qwen team terminal agent, whereas Zero lets you choose among dozens of providers |
| Grok Build | yes | 27,519 | the xAI terminal agent, tied to a single vendor |
README in short
The core idea is a terminal agent you own, with your choice of model and control over permissions. Separate sections explain the security model, the web_fetch tool, theming and contributing.