Knowledge exploration,
powered by artificial intelligence.
Latent Lab is an AI-driven graphical knowledge exploration system designed to help visualize and understand complex datasets. Explore idea spaces, simulate alternate realities with speculative AI, and extract insights from high-dimensional data.
Drive this Latent Lab from any MCP-aware host.
Drop the config into Claude Code, Claude Desktop, Cline, or the Anthropic Agent SDK. The first tool call runs a device-code login. After that every latentlab_* tool talks to this deployment.
{
"mcpServers": {
"latentlab": {
"command": "python",
"args": [
"-m",
"latentlab.mcp"
],
"env": {
"LATENTLAB_API_URL": "https://latentlabdev.media.mit.edu/api"
}
}
}
}Data Harmonization
Overlay diverse datasets, creating a unified, coherent data landscape. Align and reconcile information from varied sources for analysis and decision-making.
Explore investigations
Trend Analysis
Analyze vast amounts of data to uncover emerging patterns and trends for strategic planning and innovation.
Explore investigations
White Space Exploration
Identify gaps and uncover new areas of opportunity that were previously overlooked — surface the questions no one has asked yet.
Explore investigations
An instrument for latent knowledge.
We build cartographies of meaning — interactive maps that let researchers, designers, and strategists see what a body of text, code, or imagery is shaped like.
Latent Lab is a research project of the Viral Communications group at the MIT Media Lab, with pilots across publishing, energy, telecom, defense, and consulting.
Publications
03 · selectedHelp us upgrade the channels of thinking.
We’re looking for researchers, designers, and engineers intrigued by the future of human–AI interaction — people who want to expose latent knowledge and expand the human capacity to grasp higher dimensions of information.
Contact us