by Peiiii
An open-source, extensible, self-hosted agent workspace with multi-runtime support for Codex, Claude Code, and more, plus reusable local apps for custom interfaces and services.
# Add to your Claude Code skills
git clone https://github.com/Peiiii/nextclawLast scanned: 7/14/2026
{
"issues": [],
"status": "PASSED",
"scannedAt": "2026-07-14T06:10:37.592Z",
"npmAuditRan": false,
"pipAuditRan": true,
"promptInjectionRan": true
}nextclaw is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by Peiiii. An open-source, extensible, self-hosted agent workspace with multi-runtime support for Codex, Claude Code, and more, plus reusable local apps for custom interfaces and services. It has 249 GitHub stars.
Yes. nextclaw passed SkillsLLM's automated security scan — a dependency vulnerability audit plus prompt-injection heuristics — with no high-severity issues. You can read the full report in the Security Report section on this page.
Clone the repository with "git clone https://github.com/Peiiii/nextclaw" and add it to your Claude Code skills directory (see the Installation section above).
nextclaw is primarily written in TypeScript. It is open-source under Peiiii on GitHub, so you can review or fork the full source.
Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh nextclaw against similar tools.
No comments yet. Be the first to share your thoughts!
Get real work done with AI on the computer you control.
Tell NextClaw what you want done. It brings files, models, agents, skills, browser tools, local apps, automations, and messaging channels into one task, then keeps working toward a usable result.
Website · Download · Install · Documentation · Releases

NextClaw is a local-first AI workspace for tasks that need more than a single answer. A conversation can keep its files, references, tools, generated results, and follow-up work together instead of making you restart in separate apps.
Ready on a fresh install: NextClaw includes OpenCode Zen free-trial models, so you can start the first task without entering an API key. Free-trial access uses a public gateway; limits and models may change, and sensitive or confidential data should not be sent through it.
When scheduled work, a background Agent, or a long-running monitor finishes, NextClaw can deliver the report to your inbox. Read it later, manage it with the rest of your results, or continue the conversation with the full context.
Keep an Agent's identity, workspace, memory, and skills, then run the task with Native, Codex, Claude Code, OpenCode, or Hermes. The real task below uses Codex to refine a project architecture while the generated Markdown stays open beside the conversation.
Open code, Markdown, HTML, Word, Excel, and PowerPoint without losing the task that produced them.
Build a page with an Agent, run it beside the conversation, and keep it as a Panel App you can open and improve later.
The desktop app is the easiest way to start on macOS, Windows, or Linux.
Download the latest stable desktop release
Install Node.js LTS first, then run:
npm install -g nextclaw
nextclaw start
Open http://127.0.0.1:55667 and start with the built-in free-trial model. Connect your own provider when you need it.
If npm is unavailable, install or reinstall Node.js LTS and reopen the terminal. On a remote host, port 55667 serves plain HTTP. Use it directly only for a quick check; terminate HTTPS with Nginx or Caddy for regular access.
nextclaw stop
For a long-running server or cloud VM deployment:
curl -fsSL https://nextclaw.io/install-docker.sh | bash
See the Docker deployment guide for reverse proxy, domain, and remote access setup. You can compare every supported path on the install options page.
From the repository root:
pnpm install
pnpm dev start
The development stack prints its local URLs in the terminal and uses ~/.nextclaw by default. Set NEXTCLAW_HOME=/path/to/home to use an isolated data directory.
To run only one side:
pnpm dev:backend
pnpm dev:frontend
See the developer command reference for local source-runtime checks, the manual runtime-update harness, platform stacks, Docker, and validation commands.
To refresh the repository and website screenshot set:
pnpm run screenshots:refresh