AI Agent

Housecat is an AI agent that you can chat with to help plan projects, connect to services, read / draft / write messages, write scripts, code, and more.

How It Works

First, you need a computer to run things on. Housecat gives you a small dedicated and secure “sandbox virtual machine”.

Next, you need a Large Language Model (LLM) that receives messages (e.g. your chats) and returns intelligent responses. Housecat defaults to Anthropic’s Claude models, but Google and OpenAI models are possible too.

Additionally, you give the LLM definitions of tools it can use. When replying to your messages, the model will intelligently respond with an instruction to use tools when appropriate. Housecat defines tools to read/search/write files, read/search/write data from services like Gmail and Slack, run programs, and use a browser. These enable the system to store and retrieve text files, code, run programs, and interact with the Internet.

Now, you need an “agentic harness”. A program on the VM constantly waits for a user’s messages, then “takes a turn” and sends the instructions to the LLM API, gets a stream of messages back, runs tools and sends their results back to the LLM, and sends messages back to the user.

A VM, LLM, and a harness that continually takes turns between user messages, LLM responses, and tool calls and results can accomplish incredible things on your behalf.

Use Cases

Dig into specific use cases:

• Generate a morning briefing
• Automatically update your CRM

Further Reading

• The Unreasonable Effectiveness of an LLM Agent Loop with Tool Use
• Designing agentic loops