Toolkit

The toolkit is a Claude Code plugin: a set of skills, an orchestration agent and a knowledge base for working with worlds. Its skills follow the life of a world: structure it, grow it, play in it and keep it true, build on it. Any AI that reads markdown can use its files; it is built for Claude Code.

Repository: github.com/OnlyWorlds/toolkit

Install

In Claude Code:

/plugin marketplace add OnlyWorlds/toolkit
/plugin install toolkit@onlyworlds

Restart Claude Code to load it. Or load it straight from a clone, with no restart:

git clone https://github.com/OnlyWorlds/toolkit
claude --plugin-dir ./toolkit

Other AIs

Point the assistant at the knowledge files, in this order:

1. principles.md: the shared rules every skill follows
2. world-folder.md: the world as files, and keeping its history
3. world-memory.md: loading a world into AI sessions and writing each session back
4. onlyworlds-core.md
5. modeling-patterns.md
6. schema-reference.md
7. ai-builders.md

The skill files themselves sit in the repository’s skills/ folder, one folder per skill, and read as plain markdown.

Skills

Stage	Skill	What it does
Start	onlyworlds-start	The entry point: works out your situation and routes you to the right skill.
Structure it	parsing	Turns text, files or folders (notes, novels, campaign docs, wikis) into a structured world: JSON, or a world folder Atlas opens.
	modeling	Design help for complex systems (magic, politics, economies, tech trees) as linked elements.
	schema	Field reference across all 22 categories: look up fields, validate structure, tell similar categories apart.
Grow it	survey	Reads a world and writes a creative brief of its themes, tensions and structure.
	link	Finds missing connections, enriches links and resolves orphaned elements.
Play in it and keep it true	context	Designs how the right part of a world gets loaded into each AI session: fresh, scoped, sourced.
	resolve	Designs how a finished session (notes, recap, transcript) gets written back into the world, so canon holds.
Build on it	api	Reads, creates, updates and deletes elements through natural language, handling auth, endpoints and field mapping.
	dev	SDK integration, project scaffolding, credentials and deployment.
	council	Browse and draft schema governance motions when the 22 categories don’t fit.
	project-setup	Connects a project to an OnlyWorlds account world: credentials and a local cache. Runs when another skill needs it.

The orchestration agent, ow-agent, chains several skills for multi-step jobs, such as parsing a whole corpus and then linking and surveying it. The knowledge base covers the schema reference, the category descriptions, and decision trees for ambiguous modeling choices.

A World Folder or an Account World

Most of the toolkit runs on files alone, with no account:

• parsing, modeling and schema work fully standalone.
• survey and link read a world folder straight off disk: an Atlas world, converter output, or anything in the world folder shape.
• parsing can write that folder shape directly, so your notes become a world Atlas opens.
• context and resolve work with whatever holds the world: a folder, your own notes, or an OnlyWorlds world.

With an OnlyWorlds account, the same skills work on an account world over the API. project-setup connects the project (a .env you fill with your key and PIN as ONLYWORLDS_API_KEY and ONLYWORLDS_API_PIN, kept out of git), and api and dev build on it. To connect an agent to a world directly, see the MCP server.

Example Requests

"Help me organize my worldbuilding notes"
"Turn my notes into a world folder"
"How should I model a magic system?"
"What fields does Character have?"
"Find orphaned elements in my world"
"My AI keeps forgetting my campaign's canon"
"Turn last session's notes into world updates"
"I'm building a game with world data"