Stories
Novels and stories built from a world the AI models first, with whole lives behind every character. Or write it yourself and get feedback grounded in your world.
Let your AI build
a world to think in.
Stories, game worlds, histories and its own memory, kept as models it can come back to.
Install from meaningmodel.ai/install.md
The Meaning Model in under two minutes: a story's world, built and explored in the viewer. Watch on YouTube.
What you can build
One way of thinking, for every kind of world.
Worlds
Game worlds, histories and invented places. Sketch the whole world coarsely, then open detail wherever a reader or player goes.
Memory
An agent's memory of its work, or of the person it helps, kept as processes over time. A correction keeps what was believed before.
Real systems
Markets, institutions, technologies. Sources, estimates and uncertainty stay apart, so the model can be tested against what happens.
See it working
Worlds built with the Meaning Model.
Get started
Running in a few minutes.
Let your agent install it
Paste this into Claude Code, Codex or another agent that can run commands:
Install the Meaning Model MCP server for me by following https://meaningmodel.ai/install.mdThe install guide is written for agents: it checks requirements, installs the package and engine, connects your assistant and confirms that work is saved.
Or install it yourself
With Node.js 22.18 or newer, on macOS, Linux or Windows:
mkdir -p ~/meaning-model/data && cd ~/meaning-model
npm install @emergent-wisdom/meaning-model-mcp
npx meaning-model-mcp --install-engineConnect Claude Code
MM="$HOME/meaning-model"
claude mcp add meaning-model -s user \
-e LIFE_SIM_STATE_FILE="$MM/data/meaning-model.sqlite" \
-- "$(command -v node)" \
"$MM/node_modules/@emergent-wisdom/meaning-model-mcp/mcp-server/bin/meaning-model-mcp.mjs"Connect Claude Desktop
Open Settings → Developer → Edit Config, add this under mcpServers with your own paths (which node gives the node path), then restart Claude Desktop.
"meaning-model": {
"command": "/path/to/node",
"args": ["/Users/you/meaning-model/node_modules/@emergent-wisdom/meaning-model-mcp/mcp-server/bin/meaning-model-mcp.mjs"],
"env": { "LIFE_SIM_STATE_FILE": "/Users/you/meaning-model/data/meaning-model.sqlite" }
}Connect Codex
Add this to ~/.codex/config.toml with your own paths, then start a new session.
[mcp_servers.meaning-model]
command = "/path/to/node"
args = ["/Users/you/meaning-model/node_modules/@emergent-wisdom/meaning-model-mcp/mcp-server/bin/meaning-model-mcp.mjs"]
[mcp_servers.meaning-model.env]
LIFE_SIM_STATE_FILE = "/Users/you/meaning-model/data/meaning-model.sqlite"For stories and writing feedback, also set MEANING_MODEL_ADDONS=storytelling. Details, upgrades and removal are in the install guide.
Then ask
- “Build the world for a story about a lighthouse keeper's family, then open it in the viewer.”
- “Remember what we decided today and why. Pick it up next time.”
- “Give me feedback on this chapter.”
- “Model the coffee market over the last ten years.”
Questions
Is it free?
Yes. The code is MIT-licensed and the papers and content are CC BY 4.0. You only pay for the AI assistant you already use.
Does my work leave my computer?
The server and its database run locally. Your assistant sees what it works on, as it always does. Nothing else is sent unless you switch on the optional estimator service.
Will the AI write my book?
Only if you ask it to. You can write every word yourself and use the model for feedback, with every decision left to you.
Do I need to read the papers?
No. The tool gives your assistant its working guidance. The paper is there if you want the theory.
The Meaning Model is experimental research software. It checks and stores structure; it does not certify that a model is true, and the paper's learning claims are still untested.