Back to list
Product LaunchAI AgentsIRCSelf-Hosting

Building a Digital Doorman: Deploying an AI Agent on a $7 Monthly VPS Using IRC Transport

Developer George Larson has reimagined the traditional AI resume chatbot by building a 'digital doorman' system that operates over IRC. Moving beyond simple text rephrasing, this architecture utilizes two distinct agents—nullclaw and ironclaw—to provide deep, evidence-based answers about his work. The system is designed with a strict security boundary: a public-facing agent on a minimal $7/month VPS handles initial queries and code analysis, while a private agent on a separate machine manages sensitive data like email and calendars. By choosing IRC as the transport layer and using a lightweight Zig binary, Larson maintains full ownership of the stack while ensuring the system remains resource-efficient and aesthetically aligned with his terminal-based portfolio.

Hacker News
Share

Key Takeaways

  • Beyond Basic Chatbots: The project moves past simple RAG-based resume bots by allowing the agent to clone repositories and analyze actual code to answer technical questions.
  • Dual-Agent Architecture: A two-tier security model separates the public-facing 'nullclaw' agent from the private 'ironclaw' agent via a private IRC channel.
  • IRC as Infrastructure: The system uses IRC (via the Ergo server and Gamja web client) instead of modern APIs like Discord to ensure full stack ownership and low overhead.
  • Extreme Efficiency: The public agent is a 678 KB Zig binary that consumes only 1 MB of RAM, running on a low-cost $7/month VPS.
  • Security Boundaries: Private data (email, calendar) is isolated on a separate box accessible only through a private backoffice channel via Tailscale.

In-Depth Analysis

Rethinking the AI Portfolio Experience

George Larson identifies a common flaw in modern AI portfolio sites: they act as mere mirrors for existing resumes. To solve this, Larson developed a system where the AI doesn't just repeat facts but substantiates them. When queried about technical practices like test coverage, the agent is capable of cloning a GitHub repository, counting tests, and reading CI configurations to provide specific, data-driven responses. This shift from "parlor trick" rephrasing to active repository analysis represents a more functional use of AI agents in professional branding.

The Two-Box Security Model

The architecture is defined by a deliberate security boundary between two agents. nullclaw, the public doorman, resides on a minimal perimeter box. It is a highly optimized Zig binary designed for low resource consumption. Its role is to greet visitors and answer questions using public GitHub data. ironclaw, the private agent, sits on a more powerful, isolated system connected via Tailscale. It handles sensitive tasks involving email and calendar access. Communication between the two occurs over a private IRC channel (#backoffice), ensuring that the public-facing box never has direct access to private context or personal data.

Why IRC and Self-Hosting Matter

Choosing IRC over platforms like Discord or Telegram was a strategic decision based on three factors: aesthetics, ownership, and stability. The IRC interface complements the terminal UI of Larson's portfolio. More importantly, by using the Ergo IRC server and the Gamja web client, Larson owns the entire communication stack. This removes dependency on third-party APIs that are subject to frequent changes or restrictive terms of service, while keeping the infrastructure costs as low as $7 per month.

Industry Impact

This project demonstrates a growing trend toward "Small AI" and self-hosted agentic workflows. By proving that a functional, multi-agent system can run on a $7/month VPS with minimal RAM usage, it challenges the notion that AI agents require heavy, expensive infrastructure. Furthermore, the use of IRC as a transport layer highlights a return to robust, open protocols for internal tool communication, offering a blueprint for developers seeking to build secure, low-cost, and sovereign AI applications without relying on centralized messaging platforms.

Frequently Asked Questions

Question: Why did the developer choose IRC instead of a modern API like Discord?

IRC was chosen to maintain full ownership of the stack, avoid third-party API changes, and match the terminal-based aesthetic of the portfolio site. It allows the developer to run the entire infrastructure, including the server and client, independently.

Question: How does the system protect private data from public users?

The system uses a two-box architecture. The public-facing agent (nullclaw) has no access to private data. Complex or private inquiries are routed through a private IRC channel to a separate, more powerful machine (ironclaw) that is isolated and connected via Tailscale.

Question: What makes this agent different from a standard resume chatbot?

Unlike standard bots that only rephrase resume text, this agent can perform active tasks such as cloning GitHub repositories, analyzing codebases, and checking CI configurations to provide specific evidence for the developer's claims.

Read original article

Related News

How to Use LangSmith for Fine-Tuning Open-Source LLMs Like LLaMA2 and GPT-3.5
Product Launch

How to Use LangSmith for Fine-Tuning Open-Source LLMs Like LLaMA2 and GPT-3.5

LangChain has introduced a comprehensive guide detailing how LangSmith supports the fine-tuning and evaluation of Large Language Models (LLMs). The update focuses on enhancing dataset management, providing developers with the tools necessary to refine model performance effectively. The guide specifically highlights practical examples for fine-tuning both open-source models like LLaMA2 and proprietary models such as GPT-3.5. By integrating LangSmith into the fine-tuning workflow, users can better manage datasets and evaluate the outcomes of their training processes. This development marks a significant step in providing structured support for the lifecycle of LLM development, from data preparation to final model evaluation.

Instagram Launches First Draft Feature to Automatically Trim Reels and Highlight Key Video Moments
Product Launch

Instagram Launches First Draft Feature to Automatically Trim Reels and Highlight Key Video Moments

Instagram has introduced a new feature called "First Draft" to its Reels platform, aimed at streamlining the video editing process for creators. The tool automatically trims video clips to focus on the most important highlights, providing a foundational "starting point" for further customization. Currently rolling out to the Instagram iPhone app, First Draft is designed to reduce the manual effort required to edit raw footage into engaging short-form content. By identifying key moments automatically, the feature allows users to quickly transition from capturing footage to the final creative stages of editing. This update reflects Instagram's commitment to lowering the barrier to entry for video creation by offering automated tools that assist in the initial assembly of Reels.

Inside IBM Granite 4.2: A Technical Deep Dive into the New Era of Open-Source Reasoning and Agentic LLMs
Product Launch

Inside IBM Granite 4.2: A Technical Deep Dive into the New Era of Open-Source Reasoning and Agentic LLMs

IBM has officially unveiled Granite 4.2, a groundbreaking family of dense, decoder-only large language models (LLMs) designed specifically for enterprise-grade reasoning and agentic workflows. Released in 3B, 8B, and 30B parameter sizes under the Apache 2.0 license, these models represent a significant leap in open-source AI capabilities. Granite 4.2 is trained on approximately 15 trillion tokens using a sophisticated five-phase strategy that extends its context window to 512K tokens. A key innovation is the introduction of native reasoning—a switchable "thinking" mode that allows the models to perform step-by-step chain-of-thought deliberation. By integrating agentic reinforcement learning (RL) within real-world sandboxed environments like OpenHands and terminal interfaces, IBM has optimized the 8B and 30B versions for complex software engineering and tool-calling tasks, setting a new benchmark for open, transparent, and high-performance AI agents.