Install
$ agentstack add mcp-vaibhavpandeyvpz-zero ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
Zero
Open-source, self-hosted web UI for your own AI agent.
Chat with a fully local agent, wire up MCP servers and skills, connect channels, and tune every knob — all from a clean web interface you run yourself. Powered by Hooman and built with Next.js, React, and TypeScript.
[](https://nodejs.org/) [](https://github.com/vaibhavpandeyvpz/zero/actions/workflows/ci.yml) [](LICENSE)
Why Zero?
Zero is a web front-end for the Hooman agent toolkit. You host it on your own machine or server, bring your own LLM API keys, and keep all configuration, chat history, and data on disk under your control — nothing is sent to a hosted backend.
- Self-hosted & private — runs as a single Node process; state lives in a local directory you own.
- Bring your own model — works with Anthropic, OpenAI, Azure OpenAI, Bedrock, Google, Groq, MiniMax, Moonshot, Ollama, OpenRouter, and xAI.
- Extensible — connect MCP servers and install skills without touching code.
- Open source — MIT licensed; fork it, audit it, deploy it however you like.
Features
- Streaming chat with reasoning/thinking display, file attachments, resumable sessions, and a mobile-friendly layout.
- Tool approvals — review and approve tool calls before they run, or enable "yolo" auto-approval when you trust the agent.
- MCP servers — add, edit, remove, and reload servers over
stdio(local command), streamable HTTP, or SSE transports. - Skills — list, search, install, and remove agent skills.
- Channels — let the agent subscribe to and respond to channel messages autonomously via a background daemon.
- Deep configuration from Settings: providers & credentials, named models, web search, prompts/instructions, tools, subagents, context compaction, approvals, and channels.
Requirements
- Node.js ≥ 24 (see
enginesin [package.json](package.json)). - API access to at least one supported LLM provider (an API key, or a local Ollama instance).
The hoomanjs runtime is installed as a normal npm dependency (see [package.json](package.json)); no separate install is required.
Quick start
git clone https://github.com/vaibhavpandeyvpz/zero.git
cd zero
npm ci
npm run build
npm start
Then open http://127.0.0.1:3030.
On first run, open Settings → Providers to add an LLM provider and credentials, then Settings → Models to define a model. After that you're ready to chat.
Configuration
Data directory
Zero keeps all of its state in a single directory (created automatically):
- Default:
~/.zero - Override with the
ZERO_HOMEenvironment variable.
That directory holds:
| Path | Purpose | | ----------------- | ------------------------------------------------------------------------------ | | config.json | Agent config: providers, models, search, prompts, tools, compaction, reasoning | | instructions.md | System instructions / persona editable from Settings → Prompts | | mcp.json | Configured MCP servers | | skills/ | Installed skills | | sessions/ | Saved chat sessions | | attachments/ | Uploaded chat attachments | | memory/ | Agent memory, stored as JSONL files |
> Zero points Hooman's HOOMAN_HOME at this same directory, so everything the agent reads and writes stays under ZERO_HOME.
Environment variables
| Variable | Default | Description | | ----------- | ---------------------------------------------------------------------------------- | ---------------------------------------- | | PORT | 3030 via the dev/start scripts (3000 if you run the server entry directly) | Port to listen on | | HOST | 127.0.0.1 | Interface to bind | | ZERO_HOME | ~/.zero | Base directory for all Zero/Hooman state |
Usage
The UI has two top-level areas:
- Chat — talk to the agent, attach files, resume previous sessions, and approve pending tool calls.
- Settings — everything else, organized into tabs:
- General — agent name and basics
- Providers — LLM provider credentials/connection details
- Models — named models that reference a provider
- Search — web search configuration
- Prompts — edit
instructions.md - Tools — enable/disable built-in tools
- Subagents — configure delegated agents
- Compaction — control conversation compaction
- Approvals — manage the tool approval policy
- Channels — toggle the channel daemon so the agent can handle channel messages
Development
npm ci
npm run dev
npm run dev runs the server with tsx under pm2-runtime in development mode on http://127.0.0.1:3030 with hot reloading via Next.js.
Useful scripts:
| Script | Description | | ------------------- | ---------------------------------------------------- | | npm run dev | Start the dev server (PORT=3030) | | npm run build | Build the Next.js app and compile the server (tsc) | | npm start | Run the built server (PORT=3030) | | npm run typecheck | Type-check the app and server |
The stack is Next.js + React 19, an Express 5 API layer, Tailwind CSS with Radix/shadcn UI components, and the hoomanjs agent runtime.
Deployment / hosting
Zero is a standard Node server managed by PM2 (via pm2-runtime), which makes it straightforward to run in containers or on a VPS.
npm ci
npm run build
# Bind to all interfaces behind a reverse proxy:
HOST=0.0.0.0 PORT=3030 npm start
Recommendations:
- Put Zero behind a reverse proxy (nginx, Caddy, Traefik) that terminates TLS and adds authentication — Zero has no built-in auth and binds to
127.0.0.1by default. - Persist the
ZERO_HOMEdirectory (e.g. a mounted volume) so config, sessions, and memory survive restarts. - Provider API keys are stored in
config.jsonunderZERO_HOME; keep that directory secure and out of version control.
Related
Prefer a terminal-first agent with the same runtime, MCPs, skills, and channels? Check out Hooman — the CLI/ACP/MCP agent toolkit Zero is built on.
Contributing
Contributions are welcome! Please read [CONTRIBUTING.md](CONTRIBUTING.md) and [CODEOFCONDUCT.md](CODEOFCONDUCT.md) before opening issues or pull requests.
Security
Found a vulnerability? Please report it responsibly — see [SECURITY.md](SECURITY.md).
License
[MIT](LICENSE) © [Vaibhav Pandey](mailto:contact@vaibhavpandey.com)
Source & license
This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: vaibhavpandeyvpz
- Source: vaibhavpandeyvpz/zero
- License: MIT
- Homepage: https://vaibhavpandey.com/
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.