Packages

Open-source local AI agent framework — Queen, Workers, Quorum. Experimental research tool.

InkOS is a Node.js/TypeScript CLI/agent tool that autonomously generates, audits, and revises novels chapter-by-chapter using a multi-agent pipeline (plan/compose/draft/audit/revise), with continuity tracking via “truth” files, schema-validated JSON state deltas, and rollback/versioning. It also includes an optional local web studio and an OpenClaw-compatible skill interface with structured (e.g., --json) command outputs.

[English](./README_EN.md) | 中文

<div align="center">

A command-line interface for [Basecamp](https://basecamp.com).

**Unified deep-integration control for self-hosted media automation stacks.**

<div align="center">

ClawX - Graphical AI Assistant based on OpenClaw

Multi-agent AI coding assistant. Claude Code, Codex, Gemini, and GitHub Copilot brainstorm together, debate approaches, and deliver superior solutions. Like Cursor & Copilot, but with collaborative AI agents.

<p align="center">

[](https://github.com/ubgb/undersheet/stargazers)

<p align="center">

mcp-context-forge (ContextForge) is an open-source registry and proxy/gateway that federates Model Context Protocol (MCP) servers, A2A agent-to-agent integrations, and REST/gRPC APIs behind a unified endpoint. It provides centralized discovery/management, rate limiting and authentication middleware, retry policies, translation/virtualization of non-MCP services into MCP-compatible servers/tools, and OpenTelemetry-based observability.

Mission Control is an open-source, self-hosted AI agent orchestration dashboard that provides task dispatch and multi-agent workflow management, real-time monitoring (WebSocket/SSE), cost/token tracking, and security governance (RBAC, trust scoring, secret detection, and MCP/tool call auditing). It uses a SQLite-backed local-first setup (Next.js/TypeScript).

A Python SDK that enables applications to interact with Claude Code via an async query API and an interactive ClaudeSDKClient. It supports tool use, including in-process custom tools exposed through MCP servers, and hook-based control of the agent loop.

Provides Claude Code skill definitions (and an underlying Python package) to generate publication-quality academic diagrams, statistical plots, and slide decks from text or structured data, using a multi-agent pipeline with evaluation/self-critique and provider fallback across multiple LLM/VLM/image providers.

Provides a Claude Code “skill” (and a standalone Python script) that takes a YouTube URL, fetches captions/transcript and video metadata (optionally chapters/description/duration via yt-dlp), and outputs an Obsidian-ready Markdown file with YAML frontmatter (plus optional JSON/SRT exports).

ClawReins is the intervention layer for OpenClaw.

Smart LLM router — save 85% on inference costs. 46+ models, one wallet, x402 micropayments.