WillDeep CLI
A Rust agent for long tasks: background Runtime, subagents in Git worktrees, exit-code verification and a change audit export.
- 181 ★ on the repository
- Last commit
Broad access
Why this level
- The agent runs shell commands and edits files in the workspace, with write mode enabled and no approvals.
- The background Runtime stays on and worker agents write in separate Git worktrees.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/willdeepDetects 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 runs shell commands and edits files in the workspace, with write mode enabled and no approvals.
- The background Runtime stays on and worker agents write in separate Git worktrees.
This is third-party code. Review the repository files before installing.
What it does
WillDeep is a console coding agent with a persistent background Runtime. A task keeps going after the terminal closes, SSH drops or a page reloads, and the client later resumes from an event cursor. One `willdeep` binary provides a terminal interface (TUI), a web interface at `127.0.0.1:9847` and a headless `willdeep run` mode for CI. A main agent coordinates and worker agents take narrow tasks: reading, implementing, testing, running operations, judging and a deep mode. A subagent writes in its own Git worktree while the Runtime itself runs the verification command, with the exit code as the only judge. Deleting tests to turn a run green counts as cheating. Access has five gates: workspace policy, static rules, an AI judge, an OS-level fence (Seatbelt or bubblewrap) and user hooks. Every change is tied to a turn, an agent and a tool call, and `willdeep audit export` writes a report. Models connect over OpenAI Chat Completions, Responses and Anthropic Messages formats.
Who it is for. Teams in a closed environment who run agents on their own GPUs and open models and must be able to explain what an agent changed and why.
Good fit when
- A task runs for half an hour and the terminal cannot stay open or be interrupted.
- A small local model makes an edit and the test exit code verifies the result.
- A security team needs a report of who and which tool call changed a line.
- Approvals are easier to give from a phone while the agent runs on a server.
Not a fit when
- You need Computer Use and Browser Use: the author lists them as not covered.
- You need a multi-user web service.
- You want the simplest terminal agent with no background process.
Example request
Run willdeep in this repository and fix the failing tests, with the system itself running the verification command, then show the change report.Limitations
The project is young: installation comes from the `@rc` channel, and docs and the README are mostly in Chinese. Node.js 18 or newer is required for the installer, or use a prebuilt binary for macOS, Linux amd64 and arm64. The model key comes from an environment variable (`SOMEIM_API_KEY` or `ANTHROPIC_API_KEY`) or a config file, and without it only a local endpoint works. The background Runtime stays on and must be considered during updates (`daemon upgrade`). The WillDeep name and trademark are outside the Apache-2.0 code license. Write mode without approvals raises risk, and the OS fence is on only on machines that support it.
How to disable. Remove the package with `npm uninstall -g @niuwoai/willdeep`, stop the background Runtime and delete the settings from your home directory by hand.
FAQ
How do I install WillDeep?
Run `npm install -g @niuwoai/willdeep@rc` with Node.js 18 or newer. The installer downloads the GitHub release and verifies SHA256. Prebuilt archives for macOS and Linux also exist.
How do I run it in CI?
With `willdeep run --output json "task"`: the headless mode gives stable exit codes.
What does `willdeep doctor` check?
Configuration, fences and Runtime state. No provider is contacted.
How do I open the web interface and phone?
`willdeep web` starts or reuses the Runtime and opens the browser, while `willdeep qr` shows a QR code to pair a phone.
How do I roll back agent actions?
With `/rewind` in the session: it returns the dialog and files to a chosen step. The rollback itself can also be undone.
Discussion
No questions or reviews yet.
Related
CLI toolsA terminal coding agent tuned for the deepseek-v4 model: deep thinking, reasoning effort control, Agent Skills and MCP
A Python terminal coding agent in the manner of Claude Code: reads the repository, edits files, runs commands, with memory, sub-agents and MCP
Terminal coding agent in Go for DeepSeek, Kimi, GLM and OpenAI, built around prefix caching and confirmed writes and commands
A terminal coding agent in a single Go binary for DeepSeek and other OpenAI-compatible models, with a code graph and local OCR
DeepTide is a coding agent tuned for DeepSeek: a native macOS app, a Bun terminal CLI, a Rust version with a GUI and a set of skills in the skills folder
A coding agent for the terminal and desktop: it does not count "done" without tests or a diff and works with DeepSeek, GLM, Kimi, Claude, Codex and Gemini
README in short
The description covers a persistent Runtime, access from terminal, web and phone, five security gates, exit-code verification and audit. It continues with a quick start, a table of who it is for and how it differs from other agents, the limits of use and building from Rust sources.