MAX_AUTHORIZATION_SANDBOX

Local reproducible sandbox for testing signed authorization and default-deny enforcement.

brewery

A macOS app that updates Homebrew with one button

agentic-cuda-optimizer

An experimental optimization tool that finds CUDA kernel candidates through repeated generation, validation and performance measurement.

For developers experimenting with CUDA kernels for specific GPU tasks, it creates implementation candidates from task descriptions, measures their correctness and latency, and keeps the fastest validated candidate.

minienv

Deploy a remote environment for collaboration using your existing Docker Compose configuration.

minienv reads a Docker Compose configuration and a small extension to deploy services to a remote environment. If needed, you can open public endpoints with authentication to share work in progress with others.

timebar

A Mac timer that shows time remaining as a thin line below the menu bar

For Mac users who want to check the time without interrupting their workflow, it offers a shrinking progress line beneath the menu bar and a lightweight timer that stores data locally.

headwire

An experimental WireGuard mesh tool combining NAT traversal and DERP relaying

For operators connecting WireGuard peers across networks where direct connections are difficult, it provides a static mesh configuration that starts through a DERP relay and switches to a direct path when possible.

rig

Open-source cloud desktops for AI agents.

seaquel

A free embed that turns the SQL in your blog posts into real Postgres queries readers can edit and run without leaving the page.

Canary

Canary is the independent tester for code your agents write.

treepeat

Code similarity detection using treesitter.

Critic

A narrative viewer for reviewing coding agent changes in context

Developers reviewing agent generated code can ask the authoring agent questions on the changes page and continue the answer workflow with the original session context intact even if a device disconnects.

malkuth

A small multilingual judgment model focused on Korean that returns probabilities for fixed choices

A judgment model in 2B and 4B sizes, designed for developers who need multilingual classification with request specific criteria and probabilities for fixed choices instead of generated responses.

todo-flow

Pick TODOs. Let replaceable agents carry them through parallel work, review and verified delivery.

jevcompat

A testable spec and conformance suite for Jev-compatible (TypeSafe System One) API servers — with a normalising proxy, a reference mock and a GitHub Action.

jev-effort

Per-step reasoning effort for Claude Code, chosen by Jev, without breaking the prompt cache.

fituna

Stop guessing your llama.cpp config.

Carbon Manufacturing Systems

Own the stack: run Carbon on your own infrastructure — on-prem, in your VPC, or fully air-gapped.

Tachyon MCP is a production Model Context Protocol server runtime for Java and Kotlin.

Simten

Light the led unless both switches are on.

Crest

When Claude Code needs a permission or asks a question, it shows up on the notch: Allow, Deny, or pick an option, and the reply lands straight in your terminal.

agentrun

A workflow language for defining an existing agent's work as repeatable steps and decisions.

Lets developers who have connected agents to an application define steps and decision criteria, then call agents when needed to run workflows consistently.