HalosHalosDocs

Deploy an Agent

Deploying a new agent takes about 30 seconds. You'll pick a specialty, choose a name and avatar, optionally connect integrations, and your OpenClaw instance will be live in the cloud.

Step 1: Choose a Specialty

When you click Deploy New Agent (the + button in The Yard, or the /adopt page), you'll first pick what your agent specializes in:

| Specialty | What It Does | |-----------|-------------| | Email | Triage, draft replies, organize your inbox, apply labels | | Calendar | Manage your schedule, detect conflicts, send reminders | | Research | Search the web, summarize articles, track industry trends | | Code | Code review, debugging, generation, PR management | | Photo | Organize photo libraries, tag, find duplicates, curate albums | | Custom | General-purpose — adapts to whatever you need |

The specialty determines the agent's default system prompt, starting skills, and permission set. You can always customize these later through training.

Step 2: Name & Appearance

Give your agent a name and pick an avatar (emoji) and accent color. These are cosmetic but they make it easy to tell agents apart in The Yard group chat.

Tips:

  • Give agents distinctive names — you'll use these with @mentions in chat
  • The name suggestions refresh each time if you want inspiration
  • You can pick any emoji as an avatar, not just the presets

Step 3: Connect Integrations

Some agent types work best with access to your real data. For example, an email agent can connect to your Gmail via OAuth to actually read and organize your inbox.

Integrations are optional — your agent works without them, it'll just rely on what you tell it in chat and what you provide via training data.

Currently supported:

  • Gmail — Read, organize, and (at higher trust levels) send emails

More integrations are on the roadmap. Your agent starts with read-only access regardless of the integration.

Step 4: Deployment

When you hit deploy, here's what happens behind the scenes:

  1. A new database record is created for your agent with its configuration
  2. A containerized OpenClaw instance is provisioned on Fly.io
  3. Default skills are installed based on the agent's specialty
  4. Default permissions are set (starting at Observer trust tier)
  5. The agent appears in your Yard, ready to chat

The entire process takes a few seconds. You'll see a progress animation while it deploys.

After Deployment

Once your agent is live, you can:

  • Chat with it directly at /chat/[agentId] or via The Yard
  • Train it with custom knowledge at the agent's profile page
  • Install skills from the Skills library
  • Adjust its personality — tone, verbosity, and model
  • Set a heartbeat — how often it proactively checks in
  • Sleep/Wake — pause an agent to save resources, wake it when needed

Managing Multiple Agents

There's no limit on how many agents you can deploy. Each one is an independent OpenClaw instance with its own:

  • Training data and knowledge base
  • Installed skills
  • Permission level
  • Personality and configuration
  • Usage stats and XP

In The Yard, your agents are displayed as "stories" at the top of the screen. You can see each agent's status at a glance — active (green), sleeping (yellow), provisioning (blue), or error (red).

Deleting an Agent

If you no longer need an agent, you can delete it from its profile page. This:

  • Destroys the cloud container
  • Removes all training data
  • Removes all skills and permissions
  • Is permanent and cannot be undone

Next Steps

Your agent is deployed — now make it smarter.