physical-ai-safety-eval

A safety evaluation harness for checking the intentions of AI agents that operate physical equipment

A tool for teams evaluating whether actions proposed by AI agents operating laboratory or factory equipment are permitted under declared hardware and deployment conditions.

weir

A CI gate that checks AI agent traces for improper flows of sensitive information.

A tool for development and security teams that want to use OpenTelemetry traces already emitted by AI agents to check in CI whether sensitive data reached an unauthorized destination.

watermarks-remover

A tool for cleaning AI watermarks from content you own with privacy in mind.

An agent skill and Python service for content owners who want to remove provenance markings from multiple providers in AI-generated text and files.

Build your own theme park

Create your own theme park in seconds. Describe the roller coasters, rides, attractions and scenery you want, and watch your park come to life.

tare

A diagnostic tool that tracks Claude Code token usage from local session logs

A tool for users who want to understand why they reached a Claude Code usage limit and where tokens went by project and model. It reads session logs already on the computer to answer natural-language questions.

Anything Piano

An iOS learning app that turns a song you choose into an AI piano-roll tutorial.

Paste a song link or upload a file, and AI creates a piano tutorial in about 1 minute so beginners can practice by following falling notes.

OpenLayer

An open-source Photoshop plugin that imports ComfyUI results as editable layers

An open-source UXP plugin for image creators who want to connect a local ComfyUI server to Adobe Photoshop, preview generated or transformed results, and import them as editable layers.

experiential

An open-source gateway connecting different AI models through one control plane

It provides a single control plane and an OpenAI-compatible gateway for agent workflows that use proprietary, open-source, local, and custom models together.

conductai

A governance tool that applies policies before AI agents call models or tools and audits every decision

It applies shared policies to LLM, shell tool, and MCP calls for development and security teams operating AI agents. Before sending a request to a provider, Guard checks policy and budget and records decisions such as blocks, warnings, and audits.

rundown

An open-source client for summarizing long Hacker News discussions and exploring them through chat

A desktop client that summarizes long Hacker News posts and comment threads and lets users chat about an article or points raised in its discussion.

Termux-Dev

A terminal AI coding agent for Android Termux and desktop systems

A terminal agent for developers who want autonomous AI pair programming on Android Termux or a desktop. It offers Plan and Agent modes, multimodal vision, a live web server, diagnostics, and rollback.

Harden

An AI safety guardrail that checks coding agent actions locally before execution

It analyzes tool calls locally before execution so development and security teams using coding agents can control risky actions in advance. It also checks software before release.

singular-lite

An orchestration engine for coordinating autonomous AI coding agents in parallel within a repository

A Bash and Python engine for teams running multiple autonomous AI coding agents in a software repository. It coordinates parallel work with 3-level scheduling, persistent leases, an audit pipeline, and Git worktree isolation.

Leadcode

A tool that separates development accounts for multiple clients into macOS workspaces.

Agencies and contractors working with multiple companies can connect GitHub, Claude Code, and Codex accounts to client-specific macOS workspaces, keeping development environments separate without logging out or switching authentication.

An AI monitoring tool that detects and reports meaningful webpage changes

Users who need to track webpage changes can set up monitoring in natural language and receive a real-time change feed with fewer false positives and immediate alerts.

understudy

A test framework that checks AI agent behavior with realistic scenarios.

A scenario-based test framework for developers who want to simulate multistep behavior of conversational and autonomous AI agents and evaluate it through execution traces centered on tool calls.

proval

A self-hosted LLM agent that reviews PRs and issues on Git hosts using your own infrastructure.

Development teams that want to use their own models and infrastructure without an external review SaaS can review pull requests and issues on GitHub, GitLab, and Forgejo with an LLM.

hrdx

A lightweight multiplexer for managing multiple projects and AI agent terminals side by side

An experimental lightweight terminal multiplexer for developers who want to run AI agents such as Codex CLI and Claude Code alongside shells across multiple projects.

beckon

A tool that signals Claude Code status with distinct sounds and sends terminal notifications over SSH

For Claude Code users who cannot keep watching the terminal, it uses different sounds to signal when an agent finishes, needs a decision, or fails. When running remotely over SSH, it uses the terminal bell and desktop notifications.

shadok-ai

A local agent cockpit for managing multiple Claude Code tasks with schedules and worktrees

A tool for developers who want to run Claude Code sessions in parallel in separate Git worktrees and view or answer questions and status changes in a browser or Telegram.

pi-black

An unofficial package that connects an existing Claude subscription to Pi

A Pi package that helps developers use a Claude Max or Pro subscription in Pi by routing Anthropic OAuth requests through their existing subscription usage.