vvvkernel-cli

Command-line interface for VVVKernel — the Venice AI substrate terminal for planning and executing token launches on Base via Flaunch.gg.

Chat with 7 expert roles, run prebuilt skills, or expose the kernel to Claude Desktop / Cursor over MCP — all from one binary.

$ npx vvvkernel-cli chat "launch playbook for first 24h" --expert="Brand Expert"

Install

npm install -g vvvkernel-cli

Or run without installing:

npx vvvkernel-cli <command>

Verify:

vvvkernel --version
vvvkernel manifest

Quickstart

vvvkernel manifest
vvvkernel experts
vvvkernel chat "bootstrap community day 1?" --expert="Community Expert"
vvvkernel plan "launch ERC20 via Flaunch.gg" --expert="Onchain Expert"
vvvkernel skill list
vvvkernel skill run launch --idea="base meme coin"
vvvkernel mcp

Commands

manifest — Fetch kernel identity, endpoints, experts, chain info.

experts — List the 7 expert roles: onchain, brand, growth, community, audit, narrative, tier-design.

chat "" [--expert=] — Short structured reply: summary, plan, risks, next_step, metric.

vvvkernel chat "bonding curve vs fixed mint?" --expert="Onchain Expert"

plan "" [--expert=] — Detailed execution plan with timeline, phases, owners, metrics.

vvvkernel plan "7-day launch runway" --expert="Growth Expert"

skill list | add | run | remove — Skill management.

Built-in skills:

| Skill | Purpose | |---------|--------------------------------------------| | launch | Full launch checklist for an idea | | holders | Holder tier design + power-user mechanics | | socials | Social copy across Twitter / Discord / TG |

mcp — Start stdio MCP server for Claude Desktop, Cursor, or any stdio MCP client.

config [key[=value]] — Read or write persisted config at ~/.vvvkernelrc.

MCP Integration

VVVKernel speaks Model Context Protocol via two transports:

| Transport | Use case | Install needed | |-----------|----------------------------------|-----------------------------| | stdio | Claude Desktop, Cursor (local) | npm i -g vvvkernel-cli | | http | Remote / direct JSON-RPC | none |

Both expose the same 3 tools: kernel_chat, kernel_plan, kernel_manifest.

Claude Desktop (stdio)

Edit claude_desktop_config.json (macOS: ~/Library/Application Support/Claude/, Windows: %APPDATA%\Claude):

{
  "mcpServers": {
    "vvvkernel": {
      "command": "npx",
      "args": ["vvvkernel-cli", "mcp"]
    }
  }
}

Restart Claude Desktop — kernel tools appear in the tools menu.

Cursor (stdio)

Settings → MCP → Add New Server:

{
  "vvvkernel": {
    "command": "npx",
    "args": ["vvvkernel-cli", "mcp"]
  }
}

Remote HTTP MCP (no install)

{
  "mcpServers": {
    "vvvkernel": {
      "url": "https://vvvkernel.com/mcp",
      "transport": "http"
    }
  }
}

Direct JSON-RPC (curl)

curl -sX POST https://vvvkernel.com/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Configuration

| Variable | Default | Purpose | |----------------|--------------------------|-------------------------------------------| | VVVKERNEL_API | https://vvvkernel.com | Kernel base URL | | VVVKERNEL_ID | vvvkernel-cli | Sent as X-Agent-Id header for auditing | | VVVKERNEL_KEY | (unset) | Bearer token (optional) |

Persist to ~/.vvvkernelrc:

vvvkernel config VVVKERNEL_API=https://vvvkernel.com
vvvkernel config VVVKERNEL_ID=my-orchestrator

Writing Custom Skills

Skills are Node modules under ~/.vvvkernel/skills/. Each exports run(args) returning a prompt:

module.exports = {
  name: 'mytoken',
  description: 'Draft a token policy for a given theme',
  async run(args) {
    const theme = args.theme || 'generic meme';
    return {
      expert_role: 'Tier Design Expert',
      prompt: 'Design a 3-tier holder system for a ' + theme + ' token. Specify tier names, USD thresholds, gated perks.',
    };
  },
};

Install and run:

cp mytoken.js ~/.vvvkernel/skills/
vvvkernel skill list
vvvkernel skill run mytoken --theme="cyberpunk terminal"

Troubleshooting

| Symptom | Fix | |-----------------------------------|-------------------------------------------------------| | ECONNREFUSED | Kernel offline — curl https://vvvkernel.com/api/health | | rate_limit (429) | Back off 30s or set a unique VVVKERNEL_ID | | upstream_5xx | Retry — upstream LLM hiccup | | MCP tools not in Claude | Fully restart Claude. Validate config JSON. | | command not found: vvvkernel | npm i -g vvvkernel-cli, or use npx vvvkernel-cli |

Links

• Site: https://vvvkernel.com
• Docs: https://vvvkernel.com/docs
• MCP: https://vvvkernel.com/mcp
• Manifest: https://vvvkernel.com/api/agent/manifest
• Twitter: https://x.com/VeniceKernel
• npm: https://www.npmjs.com/package/vvvkernel-cli

License

MIT © VVVKernel