Goose: an open source AI agent
summary
goose is an open‑source agent that runs locally and connects any LLM (Anthropic OpenAI Google or locally hosted Olama) to real systems via the Model Context Protocol so it can runs shell commands and edits files and call APIs like Jira GitHub Slack and Figma. demos show it generating a 1000+ line PR written by AI, looping on errors, compiling and running tests, processing a 100,000‑card parquet dataset, and producing a ~500‑line Pygame prototype that actually works.
AI-written summary · about
chapters (14)
- 0:00Goose AI agent project
- 1:24AI agents and Goose core loop
- 4:15Capabilities and model improvements
- 5:16Goose for developer productivity
- 6:59Figma integration use case
- 7:43General computing automation
- 8:26Live demo of Goose
- 11:55Why Goose is open source
- 13:40Model Context Protocol growth
- 16:28MCP extensions security
- 17:53Goose architecture
- 19:54Customizing Goose with tools
- 20:37CI and automation
- 21:57Getting started with Goose
related talks
12mCodename Goose: An open source local AI agent built on MCP
27mThe accidental AI revolution
37mHow We Red-Teamed Our Own AI Agent: Lessons from... (Josiah Peedikayil, HS)
45mPowering AI Agents with Temporal and Kgoose
40mAgents in Action: Advancing Open Source AI Missions
38mGenerative AI Agents: How they Work and How to Build One
30mMind the Gap - A Developer's Roadmap To Building AI Agents
24mBuilding an AI Native Engineering Org