vvvkernel-cli

One binary. Talk to the kernel, run skills, bridge to MCP.

node >= 18 size 7.2 KB deps 0 license MIT

Install

# global install (recommended)
npm install -g https://vvvkernel.com/cli/vvvkernel-cli-1.0.0.tgz

# verify
vvvkernel --version

Need a local install instead?

npm install https://vvvkernel.com/cli/vvvkernel-cli-1.0.0.tgz
npx vvvkernel --help

Configure

# point at a different kernel (default: https://vvvkernel.com)
export VVVKERNEL_API=http://127.0.0.1:8080
export VVVKERNEL_ID=my-orchestrator
export VVVKERNEL_KEY=sk-...

Or persist to ~/.vvvkernelrc:

vvvkernel config VVVKERNEL_API=https://vvvkernel.com
vvvkernel config

Commands

vvvkernel manifestIdentity, expert roles, endpoint map.

vvvkernel expertsList expert roles only.

vvvkernel chat "..."Short prompt → structured reply.

vvvkernel plan "..."Prompt → full launch plan.

vvvkernel skill listBuilt-in + installed skills.

vvvkernel skill add <name>Install a skill from the registry.

vvvkernel skill run <name>Execute a skill against the kernel.

vvvkernel mcpStdio MCP server for Claude Desktop / Cursor.

Quick try

# list built-in skills
vvvkernel skill list

# draft a launch plan as the Onchain Expert
vvvkernel skill run launch --idea="VVV on Base" --chain=base-8453

# draft a 72h social rollout
vvvkernel skill run socials --audience="base-native traders"

# talk to the kernel directly
vvvkernel chat "Lead brand or tech?" --expert="Brand Expert"

Wire into Claude Desktop / Cursor

The mcp subcommand starts a stdio MCP server that exposes kernel.manifest, kernel.chat, kernel.plan as tools.

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "vvvkernel": {
      "command": "vvvkernel",
      "args": ["mcp"],
      "env": { "VVVKERNEL_API": "https://vvvkernel.com" }
    }
  }
}

Restart the client. The kernel shows up as a tool group.

Write your own skill

Drop a JS file into ~/.vvvkernel/skills/. The default export is an object with a run({ api, args, flags, log }) function.

// ~/.vvvkernel/skills/my-skill.js
module.exports = {
  name: 'my-skill',
  async run({ api, flags, log }) {
    log('hello from my-skill');
    return api.chat({
      prompt: flags.prompt || 'Default prompt',
      expert_role: 'Growth Expert'
    });
  }
};

Run with: vvvkernel skill run my-skill --prompt="..."