Goose and GooseBot: local agents and team workflows

Goose contributor and maintainer Abhijay Jain introduces Goose as an open source AI agent that runs on a user's machine and accepts their own provider keys. He describes support for local models alongside remote providers, so running the agent locally does not by itself mean that model requests stay on the machine. The talk also names macOS, Linux and Windows as supported platforms.

The extension overview explains how MCP connects Goose to tools and applications. Jain describes MCPUI through an interactive weather example and MCPF as a way to import, extend and export MCP configurations for workspaces. Recipes bundle prompts, settings and extensions into reusable workflows. His distinction is that recipes specify what to do, while skills explain how to do it.

The ecosystem section introduces Mesh LLM for shared compute and larger models. Jain describes Goose Pond as a private smart home assistant whose voice processing also stays local. Goose Distributions covers SDK work for developers building on Goose.

GooseBot applies a shared Goose instance to team discussions. Jain presents its GitHub integration as a coding assistant that reviews pull requests, answers follow-up questions and can create fix PRs. GitLab and Slack are future targets in this talk. He says the team already uses Discord internally, but workspace setup remains difficult and needs simplification.