Deploy Agent Orchestrator — Parallel AI Coding Agents — to Your Own Server

Agent Orchestrator manages fleets of AI coding agents working in parallel on your codebase. Each agent gets its own git worktree, its own branch, and its own PR. When CI fails, the agent fixes it. When reviewers leave comments, the agent addresses them. You only get pulled in when human judgment is needed.

Setting it up manually means cloning the repo, running setup scripts, configuring YAML files, connecting GitHub tokens, and managing tmux sessions. With HowToDeploy, you get a fully running orchestrator with its web dashboard in minutes.

Why Agent Orchestrator?

• Parallel agents — spawn dozens of coding agents across different issues, branches, and PRs simultaneously
• Agent-agnostic — works with Claude Code, Codex, and Aider out of the box
• Git worktree isolation — each agent works in its own worktree, so there are no merge conflicts between agents
• Autonomous CI fixing — when CI fails, the agent reads the logs and fixes the issue automatically
• Review handling — agents address reviewer comments without manual intervention
• Web dashboard — monitor all sessions, agents, and PRs from a single UI
• Plugin architecture — swap runtimes, agents, trackers, and notifiers with 8 pluggable slots

Prerequisites

Before you start, you'll need:

• A HowToDeploy account (sign up free)
• A cloud provider API key (DigitalOcean, Hetzner, Vultr, Linode, or AWS)
• A GitHub Personal Access Token (for PR, issue, and CI integration)

Step 1: Connect your cloud provider

Go to Settings → Cloud Providers and paste your API key.

Tip: Agent Orchestrator benefits from more CPU and RAM when running many agents in parallel. The default 4GB RAM / 2 CPU handles 5-10 concurrent agents comfortably.

Step 2: Deploy Agent Orchestrator

Head to the Dashboard and find Agent Orchestrator in the AI Agents section. Click the card to open the deploy form.

You need one field:

• GitHub Personal Access Token — used for GitHub integration (issues, PRs, CI status)

Server size, region, and Node.js setup are all pre-configured.

Step 3: Configure your AI provider (optional)

Expand Advanced Settings to add API keys for the coding agent you want to use:

• Anthropic API Key — if using Claude Code as the agent
• OpenAI API Key — if using Codex as the agent

You can also use Aider with local models — configure this via the YAML config file after deployment.

Step 4: Add a custom domain (optional)

Want to access the dashboard at agents.yourcompany.com? Enter your domain in Advanced Settings.

After deployment, point an A record for your domain to the server IP, click Verify DNS, and Caddy issues the SSL certificate automatically.

Step 5: Start spawning agents

Once deployment completes, open the web dashboard and configure your first project:

1. Run ao init --auto in your project directory to generate the config
2. Use ao spawn my-project 123 to assign a GitHub issue to an agent
3. Watch the agent create a branch, write code, push a PR, and handle CI

The dashboard at port 3000 shows all active sessions, their status, and linked PRs.

How the reaction system works

Agent Orchestrator's killer feature is its reaction system:

This means you can ao spawn an issue and walk away — the agent handles the full lifecycle.

What's included

Every Agent Orchestrator deployment includes:

• Node.js 20+ — runtime pre-installed and configured
• tmux — default agent runtime for session management
• Git 2.25+ — with worktree support for agent isolation
• gh CLI — GitHub CLI for seamless GitHub integration
• Web dashboard — monitor and control all agents from your browser
• Caddy — automatic HTTPS with Let's Encrypt (when using a custom domain)
• Full SSH access — your server, your agents, your rules

Pricing

You pay your cloud provider directly for the server (typically $12-24/month for the recommended spec). HowToDeploy charges a small monthly management fee for monitoring and support.

Start with a 7-day free trial — no credit card required.

---

Ready to orchestrate parallel coding agents? Deploy Agent Orchestrator now →