Agentic App
One place to work with agents on your own servers. Follow their progress, steer the work, and keep your data on your server.
Made with curiosity.
Apps, games, tools, and experiments.
A collection of what we’ve built along the way.
Newest first
One place to work with agents on your own servers. Follow their progress, steer the work, and keep your data on your server.
Flutter for Python. Build desktop interfaces with real Flutter widgets while keeping your libraries, models, and application logic in Python.
Python → MCP
async def add(a: int, b: int):
return a + b
await server.register_tool(add)Your function. An agent’s next tool.
High-performance Python MCP clients and servers, proven in production. An exceptionally simple async API for turning functions into agent tools.
A card table to share. Play the classic partnership game with friends, find a table, or enjoy a solo game across web, mobile, and desktop.
A TensorFlow implementation of R-NET, a neural reading-comprehension model. Given a passage and a question, the model identifies an answer in the text.
An experiment in learning through play. A reinforcement-learning agent takes on Flappy Bird in a browser-based game environment.
A Racket interpreter built with JavaScript and TypeScript. Explore functional programming, evaluate expressions, and run small programs in the browser.
An input-method framework built on Windows Text Services Framework. An exploration of the machinery behind text composition and candidate selection.
Make the repetitive automatic.
pip install rabbit-botA scripting engine for keyboard and mouse input. Built to automate repetitive Windows tasks, with Python and JavaScript packages and a scriptable desktop app.
Game state. Decisions. Actions.
Python · C# · JavaScript · and more
A Hearthstone bot engine for AI research. It connects game state and actions to a separate bot service, letting researchers use their preferred programming language.