Accelerating now

This repo open-sources a full Chinese textbook on AI agent engineering—complete with Markdown source, compiled PDF, and runnable Python demos for every chapter.

It breaks the vendor lock on Codex and Claude Code by translating their API calls to any LLM backend you choose.

wigolo gives AI coding agents a local, keyless web research layer that runs entirely on your machine and bills nothing per query.

LiveAgent is a local-first desktop client that gives LLMs real file-system and shell access, plus an optional gateway for remote browser control.

NitroStack treats MCP servers as first-class backends, not afterthoughts held together with glue code.

It bundles the entire AI drama workflow—script, storyboard, voice, and final cut—into a single pipeline you can host yourself.

A skill that spawns parallel reasoning processes under distorted cognitive frames, then scores and prunes them with a separate critic pass.

This repository packages the public metadata that lets MCP clients discover and query Vybe’s live Solana API.

A demo repo for running extreme-quantized language models locally without needing a research cluster.

An unofficial proxy that borrows your ChatGPT/Codex OAuth tokens to serve a local OpenAI-compatible API, bypassing API credit billing.

A library of 75 portable Markdown playbooks that teach AI coding agents to design and build interfaces with constraints instead of vibes.

Raven wraps agents in durable memory and self-refining skills so workflows survive past the chat session.

pi-web gives the pi coding agent a local browser UI so you can browse past sessions, fork conversations, and preview files instead of digging through terminal history.

Agent skills that turn your existing LLM coding assistant into a parallelized SAST scanner without extra tooling.

Flint compiles compact, human-editable chart specs into polished Vega-Lite, ECharts, or Chart.js visualizations so agents and developers can skip hand-tuning scales, axes, and spacing.

iFixAi runs up to 32 inspections against any LLM or agent and returns a letter-grade scorecard in minutes, using a separate provider as judge so the model isn't grading its own homework.

A single Zig binary with an embedded Svelte dashboard that installs, supervises, and cross-wires local AI agents, workflow engines, and tracing tools so you don't have to juggle separate terminals.

MoonshotAI built a terminal-native agent that reads code, runs shells, and watches screen recordings so you can stop writing bug reports in prose.

It exists to stop your AI gateway from quietly burning through quotas, cash, and expired OAuth tokens without leaving a paper trail.

codex-keysmith exists because copying a Markdown file and editing one TOML key is simple, but existing files, active hooks, and interrupted writes are not.