
by putervision
World Model MCP is a local MCP server from PuterVision that keeps a persistent 3D and 2D spatial model of entities for AI agents. It tracks objects, simulates movement and checks collisions without any LLM in the loop.
PluginContact for pricing
Step 1: Install Node.js 18.18 or newer.
Step 2: Run npm install -g @putervision/world-model-mcp.
Step 3: In your project directory, run world-model-mcp init to create the .world-model-mcp folder and write MCP configs for editors such as Cursor, Claude and VS Code.
Step 4: Restart your IDE or agent manager so it picks up the server.
Step 5: Have the agent call tools such as update_entity, query_entities and simulate_movement.
Step 6: Run world-model-mcp view to open the 3D scene visualizer, or world-model-mcp stats for database metrics.
No. The README says all spatial entities, relations and history stay in the local .world-model-mcp folder and that no telemetry or project data is transmitted.
Node.js 18.18 or newer. It is installed globally with npm and started with world-model-mcp init and world-model-mcp run.
The README describes 15 consolidated MCP tools covering spatial memory, simulation, goal integration, spatial specs, and multi-agent and replay features.
Yes. Running world-model-mcp view launches an interactive 3D WebGL visualizer in the browser, and world-model-mcp stats shows database metrics and permanence confidence stats.
Requirements
InfoBroker MCPInfobroker is an MCP server that puts web search, page extraction, citations and claim verification behind one set of tools. Twenty-one of its providers work without configuration.QokedasTraining-data company that turns real-world scientific signals and observations that are not normally written down into datasets for improving AI models in science.
MONETMONET is a dataset of 103.8 million English image-text pairs for training text-to-image models. It was distilled from 2.9 billion raw pairs through safety, domain and duplicate filtering.