AgentKit is a library of prebuilt agent capability for AI coding tools. It ships 108-plus skills across development, marketing, research and automation, 95-plus commands covering the software lifecycle and the marketing funnel, and 45 agents – 17 engineering, 28 marketing – alongside MCP integrations and a desktop app for managing licences and configuration. It works with Claude Code and GitHub Copilot, and is by Goon Nguyen.
Pricing is the standout: $99 for the engineering kit, $99 for marketing, $149 for the bundle, paid once, with the desktop app at $19 a year. Against a category that has settled on monthly seats for everything, a one-time purchase of configuration you then own is a genuinely different offer, and it fits what is actually being sold – these are prompts, workflows and definitions, not a hosted service with ongoing costs to cover.
Two things to weigh. The product was called Claudekit at claudekit.cc until recently, and it renamed itself to AgentKit and states plainly that it has no affiliation with Anthropic – dropping a name built on somebody else’s trademark before being made to is to its credit, and worth noting because most products in that position do not. The substantive question is durability: skills and commands are written against the conventions of tools that change quickly, so a one-time purchase carries the risk that a paid library ages into needing rework, and the counts (108 skills, 95 commands) describe volume rather than quality.
