Where an organization and its AI reason together
Mantis builds cognitive infrastructure for the AI-native enterprise: environments where people and AI work over the same data, knowledge, software, and workflows.
Thesis
Cognitive infrastructure for the AI-native enterprise
Mantis is building cognitive infrastructure for the AI-native enterprise: systems that transform an organization’s existing data, knowledge, software, and workflows into environments where humans and AI can reason and act together.
Rather than adding another chatbot or intelligence layer on top of the enterprise, Mantis embeds AI capabilities directly into the systems organizations already use, making complex information more legible, decisions more grounded, and workflows more capable.
The thesis extends Professor Manolis Kellis’s long-standing research philosophy of using computation to reveal hidden structure in complex systems, from genomics and biological circuitry to organizations themselves. Mantis is developed in his group at MIT and CSAIL, where machine learning research sits alongside interface and systems engineering.
Built at
The MIT and CSAIL names identify where Mantis is built. Neither institution endorses the product.
- Platform
- Cognitive infrastructure: spatial workspace, map engine, and agent runtime
- Focus
- Making complex information legible, decisions grounded, and workflows more capable
- Built at
- MIT CSAIL
- Designed with
- Research groups, core facilities, and data teams working inside the systems they already run
- Analysis
- Interactive analysis, with no re-indexing
- Compute
- Runs on your own hardware or on hosted workers
Use cases
Designed for real-world work at scale
Literature review
Map thousands of papers and documents at once to see the shape of a field before reading any of them.
Scientific data
Explore single-cell, molecular, and geospatial datasets as maps instead of static, disconnected plots.
Agent-assisted research
Give AI agents and automated pipelines a native way to investigate the same map your team uses.
Embedded maps
Build interactive maps, clusters, and filters directly into products and notebooks.
Traceable analysis
Resolve high-stakes questions across teams with evidence that stays linked to its source.
Connected to the sources you work in
Mantis reads open corpora, institutional archives, and private collections, and maps them into one navigable space.
Explore Mantis
Spatial analysis solutions
Literature review, scientific data, embedded maps, and agentic research built on Mantis.
Read more
Data and model integrations
The public sources, file types, model providers, and developer surfaces that connect to Mantis.
Read more
Mantis, shown rather than described
Product footage from the workspace and the talks where the thesis was laid out.
Read more
Mantis documentation
The manual: concepts, creation routes, agents, analysis, collaboration, and the extension SDK.
Read more
Documentation by topic
The manual arranged by what you want to do, from reading the map to building an extension.
Read more
Frequently asked questions
Clear answers about Mantis maps, clusters, embeddings, privacy, and deployment.
Read more