Skip to main content

Agentic State Machines

goose is replacing its classic agent loop with a re-entrant state machine where the conversation is the state and each behavior is an Operation.
Read full article
Agentic State Machines
Moving to issues as the new PRs

Moving to issues as the new PRs

We’re changing how people contribute to goose by moving design and discussion into issues before agents turn ready issues into code.
Self-Improving Agents Still Need Humans

Self-Improving Agents Still Need Humans

Benchmarks are most useful when they become bug reports. Here's how the goose team uses Terminal-bench failures, Harbor, and human judgment to improve real agent behavior.
How to Break Up with Your Agent

How to Break Up with Your Agent

ACP lets you keep your favorite editor but swap the AI agent, or keep your agent but use it from any editor. Here's what actually works today in Goose.
WebMCP for Beginners

WebMCP for Beginners

WebMCP lets websites expose structured actions that AI agents can call directly. This guide explains how it works, how it differs from MCP and browser automation, and how to build your own WebMCP-enabled site.
Order Lunch Without Leaving Your AI Agent

Order Lunch Without Leaving Your AI Agent

Use the Neighborhood extension in goose to discover nearby restaurants, browse interactive menus, and place a takeout order, all from a simple chat prompt.
One Shot Prompting is Dead

One Shot Prompting is Dead

Practical steps and mental models for building context engineered workflows instead of clever prompts.