Getting Started

docker pull node:20-alpine

# From inside any git repository
cd your-project
GOOGLE_API_KEY=your_key npx agentnetes run "add comprehensive test coverage"

npm install -g agentnetes

# Then use anywhere
GOOGLE_API_KEY=your_key agentnetes run "refactor auth module to use JWT"

search(pattern, path?, fileGlob?)  // grep -r across the repo
execute(command)                   // run any shell command in the sandbox

docker pull node:20-alpine

# ── Sandbox ───────────────────────────────────────────────────────
SIMULATION_MODE=false
SANDBOX_PROVIDER=docker

# ── Google Gemini · get free key at aistudio.google.com ───────────
GOOGLE_API_KEY=your_key_here

# ── Target repo for agents to work on ─────────────────────────────
DEMO_REPO_URL=https://github.com/expressjs/express

npm run dev

watch -n 1 docker ps

agentnetes run "<goal>"

# Examples
agentnetes run "add TypeScript strict mode and fix all type errors"
agentnetes run "write unit tests for the auth module"
agentnetes run "implement rate limiting on all API routes"
agentnetes run "add OpenTelemetry tracing throughout the app"

VERCEL_TOKEN=your_token agentnetes snapshot create

VERCEL_TOKEN=your_token agentnetes snapshot list

# Start on default port 3000
npx agentnetes serve

# Custom port
npx agentnetes serve --port 8080

export GOOGLE_API_KEY=your_key
npx agentnetes serve

# Option 1: export in your shell session
export GOOGLE_API_KEY=your_key
npx agentnetes serve

# Option 2: add to .env.local in the directory where you run serve
GOOGLE_API_KEY=your_key

# Use Docker (default for local dev)
SANDBOX_PROVIDER=docker GOOGLE_API_KEY=xxx agentnetes run "goal"

# Use local (fastest, no isolation)
SANDBOX_PROVIDER=local GOOGLE_API_KEY=xxx agentnetes run "goal"

# Use Vercel (requires VERCEL_TOKEN)
SANDBOX_PROVIDER=vercel VERCEL_TOKEN=xxx GOOGLE_API_KEY=xxx agentnetes run "goal"

# ── LLM Provider ─────────────────────────────────────────────────
GOOGLE_API_KEY=                  # Required · get free at aistudio.google.com

# ── Sandbox ───────────────────────────────────────────────────────
SANDBOX_PROVIDER=docker          # docker | local | vercel | e2b | daytona
VERCEL_TOKEN=                    # Required for vercel sandbox

# ── Models (optional overrides) ───────────────────────────────────
PLANNER_MODEL=google/gemini-2.5-pro    # Root Tech Lead model
WORKER_MODEL=google/gemini-2.5-flash   # Specialist agent model

# ── Web app / serve only ───────────────────────────────────────────
SIMULATION_MODE=false            # true = always use simulation
DEMO_REPO_URL=https://github.com/expressjs/express

cd your-node-project
GOOGLE_API_KEY=xxx npx agentnetes run "add vitest unit tests for all utility functions"

GOOGLE_API_KEY=xxx npx agentnetes run "enable strict TypeScript and fix all type errors"

GOOGLE_API_KEY=xxx npx agentnetes run "add rate limiting middleware to all Express routes using express-rate-limit"

GOOGLE_API_KEY=xxx npx agentnetes run "audit the codebase for security vulnerabilities and fix them"

GOOGLE_API_KEY=xxx npx agentnetes run "add OpenTelemetry tracing to all API routes and database queries"

import { VrlmRuntime } from 'agentnetes/runtime';
import { VrlmEventEmitter } from 'agentnetes/events';

const emitter = new VrlmEventEmitter();
emitter.on(event => {
  if (event.type === 'finding') console.log(event.data.text);
  if (event.type === 'artifact') console.log('File:', event.data.artifact.filename);
  if (event.type === 'done') console.log('Complete.');
});

const runtime = new VrlmRuntime(emitter, {
  repoUrl: 'https://github.com/your-org/your-repo',
  plannerModel: 'google/gemini-2.5-pro',
  workerModel: 'google/gemini-2.5-flash',
  maxWorkers: 4,
  maxStepsPerAgent: 20,
});

await runtime.run('add comprehensive test coverage');