Block's Goose AI Agent: Driving Cross-Functional Productivity and Trust

Key Quotes

"oh it has been absolutely amazing the keynote was great i got a shout out so that made my entire day and i must say i am quite jealous of your jacket you see me a little little embroidered angie in there it looks great on you very very nice very nice touch thank you github welcome to the github podcast a show dedicated to the topics trends stories and culture in and around the open source community on github my name is abby and i'm one of your hosts from the open source programs team here at github and we are recording live at github universe day two where i'm sitting with angie jones angie welcome thank you"

Abby, the host, opens the podcast by expressing enthusiasm for the event and welcoming Angie Jones. This sets a positive and engaging tone for the discussion, highlighting the live recording environment at GitHub Universe. Abby's mention of Angie receiving a shout-out in the keynote suggests Angie's prominence within the community.

"yeah sure um angie jones i'm vp of engineering at block focused on ai tools and enablement uh and so i mean i don't know everybody's focused on ai right now um but yeah a lot of what i do is helping developers i guess navigate this new new age of you know modern software development with ai assistance right and i do that both externally and internally with our own engineering department and then externally um i over i also own a devrel function and so um our developer advocates are outside as well like teaching developers all over the world yes"

Angie Jones introduces herself as the VP of Engineering at Block, focusing on AI tools and enablement. Angie explains her role involves guiding developers through modern software development with AI assistance, both within Block and externally through a developer relations function. This establishes her expertise and the scope of her work in the AI space.

"so goose is our open source ai agent um it actually was developed as the reference implementation of mcp with anthropic so um we were one of the launch partners for the protocol and we developed goose as basically the proof uh of of what an agent that supports mcp could look like so yeah that's goose um we we developed that internally for our own engineers and decided to open source it in january um you know after we saw that we were getting like a lot of benefit from it so we wanted to share that with the community as block does"

Angie Jones describes Goose as Block's open-source AI agent, developed as a reference implementation for the Model Context Protocol (MCP) with Anthropic. Angie explains that Goose was initially created for Block's internal engineers and later open-sourced after the company recognized its significant benefits. This highlights Goose's role as a practical demonstration of MCP capabilities.

"yeah yeah you know and that's the beauty of open source i think um by goose being like that open source client this gave the community a voice to express what do they want out of agentive ai and so goose has the most permissible license we accept contributions from the community and so they're the ones that are really kind of moving the needle on we want this we want that things like multimodal support for example you know what i mean uh like a lot of the patterns you see it first in goose which is great and then other bigger clients we see then adopt but that's perfect because this gives the community an avenue to basically show what they want and demand the features that they want out of this and then you know we see the ecosystem move accordingly yeah"

Angie Jones emphasizes the value of open source, explaining how Goose, as an open-source client, empowers the community to shape the future of agentive AI. Angie notes that Goose's permissive license encourages community contributions, driving development in areas like multimodal support. This demonstrates how community input directly influences the evolution of AI tools and features.

"so what examples have you seen of people building with goose um we see all kinds of things like people think of agents and mcp and all of that as a developer tool we are using goose and mcp across our entire company wow so we have 12 000 employees um of 15 plus different job functionalities everyone is using like multiple agents in a day you know connected via mcp to the tools that they use every day i'm talking finance marketing sales executive assistance everyone is using these agents and so i feel like i almost have a seat in the future of what this is going to look like i feel like everyone else is so focused on the developer part of this but like this is being used across the board"

Angie Jones shares that Goose and MCP are being utilized across Block's entire company, not just by developers. Angie highlights that 12,000 employees across various job functions, including finance, marketing, and sales, use multiple agents daily connected via MCP. This illustrates the broad applicability and integration of AI agents beyond traditional software development roles.

"oh everyone's like everyone has extra hands now essentially right and also like the ability to kind of multitask in ways that we've not done before like i think everyone has pretty much gotten comfortable like doing multiple things at a time but now you can do even more and like expedite that quite a bit by using agents to like set set off things for example i see engineers and they're like okay i need i'm going to write this code because i like to write code i want to write this code but while i'm doing this i'll have you know maybe an agent go and kind of run or write some tests that's what they're for this right if we both have the design document then we can work in parallel have one agent do this i'll write the code you know and and if there are any other things maybe you need to set up environments or something so you can kind of just like delegate these tasks off to these various agents"

Angie Jones explains that AI agents provide individuals with "extra hands," enabling unprecedented multitasking and expediting workflows. Angie uses the example of engineers writing code while simultaneously having an agent run tests or set up environments. This demonstrates how agents can delegate tasks, allowing humans to focus on core responsibilities and work in parallel.