Open source · Apache-2.0
neuro-kitchen is a growing toolkit for research: GPU signal processing and spike sorting, image analysis for histology, data conversion and paper management. The apps share one frame, and work with the agents you already use.
Multi-channel signal processing and spike sorting, written once and run on any GPU (Vulkan, Metal, CUDA, ROCm) or the CPU. A desktop app, Rust crates, a Python package and a command line.
pip install dspkView on GitHub A GPU workstation for histology and microscopy: whole-slide and OME-Zarr images, annotation, pipelines and measurement tables.
Coming soon
What the apps share: one look, one frame and one list of actions, so every tool works the same way, for people and for agents.
cargo add kitchen-uiView on GitHub Convert neuroscience data between formats.
View on GitHubA local-first reference manager for collecting, reading and citing papers, on desktop and mobile.
Coming soon
The apps share one look, one frame and one way to work, so learning one is learning them all.
Released
The heavy work is written once in CubeCL and runs on Vulkan, Metal, CUDA or ROCm, or on the CPU, from the app, Python or the command line.
In progress
Every action of an app is declared once, with what it does and when it can run. Menus, keyboard keys, the search and agents all use that same list, so none drifts from the others.
In progress
An app serves its actions to agents over MCP, and its chat runs the agent you already have (Antigravity, Claude Code, any ACP agent) with your own login. No API keys in the app, and every action an agent takes asks first and is recorded.
In progress
A timeline of what you and the agent did: which kind of work, when, for how long, summarised into a few episodes. Click one to see each step, where it came from and what it produced.
In progress
A project keeps its steps, its pipeline and its scripts’ locked environments (uv.lock, Cargo.lock), so the same analysis runs headless over many files, or on another computer.
01
Apache-2.0 code, public issues, and decisions made where everyone can see them.
02
Versioned, citable releases so a result in a paper can be rerun years later.
03
Designed around real lab workflows, and improved by the people who use the tools every day.
// JOIN IN
Students, researchers and engineers who want tools they can open up, understand and build on. If one of our tools saved you an afternoon, we'd love to hear about it. And if you're able to chip in, you help the toolkit keep growing.