Show HN: I put an AI agent on a $7/month VPS with IRC as its transport layer
What it is
Think of it as giving an AI agent a pager from the '90s. Instead of fancy webhooks or REST APIs, this setup uses IRC—the text-based chat protocol from 1988—to send commands to an AI agent running on a budget VPS. The agent joins IRC channels, listens for instructions, performs tasks (web searches, code execution, file operations), and posts results back to the channel.
Why it matters
This matters because most AI agent tutorials assume you're running on expensive cloud infrastructure with complex orchestration. IRC is simpler: no OAuth, no rate limits, no vendor lock-in. It's human-readable, debuggable with any IRC client, and costs almost nothing. If you're building personal automation or internal tools, this approach lets you prototype fast and keep costs flat.
Key details
- •Uses IRC (Internet Relay Chat) as the transport layer — plaintext, stateless, debuggable with any IRC client
- •Hosted on a $7/month VPS (likely a basic DigitalOcean or Linode droplet)
- •Agent named 'Nullclaw Doorman' — joins channels, parses commands, executes tasks, reports results
- •No complex API setup: commands are just IRC messages, responses are channel posts
- •Author's blog post includes implementation details and code examples