Top Picks

A CLI and Python harness that lets AI agents control a real iPhone through macOS iPhone Mirroring or an Android phone over adb, with OCR, accessibility, and input primitives for automation.

A research site for LittleLearner, a language model trained only on K–5 curriculum data to study how bounded pretraining affects capability, scaling, post-training, and in-context learning. The page includes a live chat demo, model scales, controls, and experiment summaries.

A Claude Code skill for building Apple-style Liquid Glass UI, with CSS fallback, a refraction layer, and install steps for Claude Code and manual use.

TypeScript SDK for AI agents and backends that connect to Squid Pay, letting agents propose payments or trades while the owner wallet remains the final signer.

A logical KV cache profiler and leak detector for PagedAttention engines like vLLM and SGLang, with a dashboard, live engine hook, CI checks, and Perfetto trace export.

Substructure is a cloud agent engine for building team-facing agents that connect to Slack, use MCP servers, and run custom agent loops through webhooks. The page shows configuration examples, CLI commands, local development, self-hosting, and support for BYO LLMs and AG-UI.

A Claude Code status-line tool that estimates per-turn costs, subagent spend, cache rebuilds, and usage limits from local session data.

Self-hosted voice assistant that can answer aloud, dictate text, run shell commands, control mouse and keyboard, and connect to MCP-enabled services. It can hand off heavier tasks to coding agents and run on local hardware or with hosted providers.

An open 45M-parameter agentic LLM for tiny devices, built for tool calling, device use, and structured extraction. It ships as a 14 MB binary and is positioned for phones, wearables, smart home devices, robots, and other low-RAM hardware.

A set of 20 Claude Code skills distilled from real engineering failures, each packaged as a short rule set with a pre-flight checklist and install instructions.

A macOS menu bar app that shows remaining Claude 5-hour and weekly usage limits locally, without reading tokens or calling an API.

clai is a terminal-first AI tool that pipes stdin to an LLM and writes results to stdout, so it fits naturally into shell workflows like git diffs, text files, and command pipelines.

FastAPI-based server that turns files and Python modules into REST APIs, browsable UIs, searchable resources, and an MCP endpoint.

Open-source security guidance for specific libraries and versions, aimed at developers and coding agents. The repo includes local setup steps, a web site, and an installable skill for agent workflows.

Rust building blocks for frontier OpenAI agents, with a CLI, SDK, and examples for agent sessions, event streaming, and voice-driven workflows.

A paid biotech intelligence brief on EGFR TKI resistance in NSCLC, covering osimertinib resistance mechanisms, competitive pipeline updates, and sequencing strategy. The page shows a downloadable report, pricing, publication timing, and a detailed summary of the brief’s findings.

Galda is a task-management workspace for Claude Code and Codex that keeps progress in reviewable summaries, helps manage multiple tasks in one place, and is built to run on a user's own machine without extra API costs.

Crew is a local-first collaborative IDE for people and AI agents to work together on projects. It supports multiple agents at once, shared threads, project folders, token tracking, and built-in browser tools like terminal, files, preview, web, and review.

Local Windows app that converts, sanitizes, and indexes documents before they reach an LLM. It removes sensitive data on-device, creates searchable Markdown/CSV outputs, and supports both personal and business editions.

Self-hosted OpenAI-compatible mock server for testing LLM integrations without exposing API keys. Includes a hosted instance, local setup, Docker support, streaming, and deterministic responses.

Open-source CLI that keeps a single coding session in sync across Claude Code and OpenAI Codex CLI, letting you switch between models without losing context or paying for two parallel runs.

ZeroShot is a developer tool for monitoring coding-agent sessions, turning sessions and PR reviews into reusable skills, and enforcing team conventions on every PR.

VS Code and Cursor extension that fixes bugs from editor diagnostics, explains the root cause, and can generate regression tests from the same flow.

AgentCouch is an MCP-based messaging layer that lets AI agents chat directly in shared rooms, hand off context, and keep replayable transcripts. The page includes install steps for Claude, Codex, Cursor, OpenClaw, and generic MCP clients, plus examples and a sample agent conversation.