| paperclipai/paperclipMIT-licensed, self-hosted Node.js and React control plane that orchestrates Claude Code and other agents through tasks, goals, org charts, heartbeats, budgets, approvals, skills, governed connections, plugins, workspaces, and audit history; it runs alongside Claude Code as a standalone host/orchestrator and exposes Paperclip operations through MCP. | Self-hosted multi-agent orchestration and governance control plane | Usable for self-hosted local and authenticated deployments; production operation requires security, credential, storage, database, runtime, and network planning. | Oct 5, 2026 |
| awesome-skills/first-principles-skillMIT-licensed Claude Code skill, version 0.2.0, that injects a five-phase first-principles method for defining a problem, challenging assumptions, establishing ground truths, reasoning upward, and validating a structured recommendation. | Claude Code reasoning skill | Usable with compatibility caveats | Oct 1, 2026 |
| mnemosyne-oss/mnemosyneMIT-licensed, local-first Python memory layer for AI agents that stores memories in SQLite and exposes remember, recall, and forget capabilities to Claude Code through a documented stdio MCP server; it also offers a Python SDK, CLI, optional local embeddings and consolidation, temporal triples, memory banks, import/export, and optional encrypted cross-instance sync. | Local AI-agent memory server and Python library | Usable for local Claude Code memory with caveats; production sync, remote models, sensitive-memory retention, and benchmark claims require additional validation. | Sep 30, 2026 |
| NandhaKishorM/layaApache-2.0 multilingual decision-model toolkit that provides Python and TypeScript SDKs, local CLI and HTTP serving, ONNX runtimes, framework integrations, evaluation harnesses, and an optional MCP server for fast typed choice, ordinal-score, and yes/no decisions; no native Claude Code plugin, hook, skill, or Claude-specific installation path is evidenced. | Local decision-model SDK and MCP tool server | Usable for evaluation and carefully validated deployments, but shipped calibration, multilingual task quality, option-order sensitivity, and workload-specific accuracy require local testing before automated actions. | Sep 30, 2026 |
| MadAppGang/magusMIT-licensed Claude Code plugin marketplace and companion plugin suite for software development, combining commands, agents, skills, hooks, MCP runtimes, terminal control, semantic code navigation, multi-model orchestration, project provisioning, UI review, and stack-specific guidance; several features require external binaries, provider credentials, or dependent plugins. | Claude Code plugin marketplace and development workflow suite | Mixed: core plugin packaging and documentation appear usable, while readiness varies substantially by plugin and external dependency. | Sep 19, 2026 |
| calcom/cal.diyMIT-licensed, community-maintained self-hosted scheduling platform forked from Cal.com with commercial and enterprise code removed; it includes a Next.js web app, NestJS API v2, PostgreSQL/Prisma data layer, booking embeds, calendar and conferencing integrations, Docker deployment materials, and repository-local spec and engineering-rule documents intended to guide Claude during development, but no distributable Claude Code plugin, skill, hook, MCP server, or command is evidenced. | Self-hosted scheduling application with repository-local Claude development context | Usable for advanced self-hosters and contributors; explicitly not recommended by its README for production or commercial deployment. | Sep 18, 2026 |