Quickstart

Deploy your first AI agent with ClawRun in under a minute.

Prerequisites

Deploy your first agent

Install the CLI

npm install -g clawrun

Run the deploy command

clawrun deploy

The interactive wizard will walk you through:

  • LLM provider. Select your provider and enter your API key.
  • Channels. Enable messaging channels (Telegram, Discord, Slack, WhatsApp, and more).
  • Tools. Choose which tools to install inside the sandbox.
  • Network policy. Unrestricted or restricted (allowlist specific domains).
  • State store. Connect an existing Redis/Upstash store or provision a new one.

ClawRun then scaffolds a Next.js orchestrator app, deploys it to Vercel, creates the first sandbox, and starts your agent.

Chat with your agent

Once deployed, the CLI automatically opens an interactive chat session. You can also send a one-off message anytime:

clawrun agent my-instance -m "What can you do?"

Or start an interactive session:

clawrun agent my-instance

Open the web dashboard

Every instance comes with a web-based chat interface. Open it with:

clawrun web my-instance

This generates a short-lived auth link and opens it in your browser.

What just happened?

Behind the scenes, clawrun deploy did the following:

  1. Scaffolded an orchestrator. A Next.js app with API handlers for webhooks, chat, sandbox lifecycle, and cron-based wake.
  2. Deployed to Vercel. The orchestrator runs as serverless functions on Vercel.
  3. Created a Firecracker microVM. A Vercel Sandbox with your configured network policy.
  4. Provisioned the agent. Installed ZeroClaw and your selected tools inside the sandbox.
  5. Started the sidecar. A supervisor process inside the sandbox that manages the agent daemon, reports heartbeats, and exposes a health endpoint.

The sandbox stays alive as long as there's activity. When idle, it snapshots and sleeps. The next message or scheduled task wakes it back up.

Next steps