Ai-Agent Builder is a production-grade, "backend-less" single-file web application for authoring, configuring, and packaging AI agents. It runs entirely in the browser — no server, no API keys stored, no telemetry — making it invisible to trackers and safe for enterprise environments.

Self-hosted lifecycle platform for software powered by Claude Code. Configurable pipelines, devices you control, no credentials on the server. Apache-2.0.

Multi-agent pipeline simulator — planner, researcher, coder, reviewer, synthesizer tracking quality and latency

Vibe-coded tower-defense game built by an AI agent pipeline (planner → coder → smoke-tester → reviewer) running across Claude Code, Cursor, and VS Code Copilot.

古籍识别三 Agent 流水线：OCR → 页序重排 → 错别字纠正 | Ancient Chinese text OCR pipeline: 3-agent system for classical books, ¥0.016/page.

Manifest-driven 21-stage agent pipeline for Python — 5 LLM backends (Anthropic / OpenAI / Google / vLLM / Claude Code CLI), tools, skills, memory, sandboxed CLI runs, MCP. The engine behind Geny. Apache-2.0.

Reuse Agent Skills across Cursor, Claude Code, Codex, and CodeBuddy. Build task-specific AI coding agent pipelines for one project, each with an independently selected capability set.

Your AI dev brigade for Claude Code. Multi-agent sprint pipeline with fan-out review, Bug Council, sprint learnings. Zero dependencies.

Production-grade multi-agent LLM orchestration — define agents, chain them, manage state, deploy. Lightweight alternative to LangChain/CrewAI.

Overnight AI pipeline that turns live panel transcripts into responses from ten non-human and historical voices — a provotype for the World Beautiful Business Forum's AI Democracy Marathon, Athens 2026.

Convert any site's sitemap into an Atom feed for editorial and agent pipelines

Orchestration system for long-running complex agent tasks

Adversarial review + provenance-audit pipeline for LLM-generated research. Catches fabrication, self-overconfidence, and data-layer inflation.

Governed text-to-SQL analytics copilot. Claude-powered LangGraph agent scoped to certified dbt metrics. LLM cannot invent numbers that don't exist. 100% eval accuracy.