Skip to content
Cognitive cartography
Read your data as a map.
Mantis embeds a corpus, lays it out as a navigable map of meaning, and puts agents on that map with you. This is the manual: how to make a space, how to read what you see, and how to drive it.
- Where
- mantis.csail.mit.edu
- Sign in
- Google or GitHub
- Built at
- Kellis Lab, MIT CSAIL
- Automate with
- Mantis CLI, extensions, macros
Start here
5 minutes
Get set up
Request access, sign in, and find your way around the home dashboard.
Create your first space
Turn a CSV, a folder of text, a paper search, or a repository into a map.
Read the map
What a point, a cluster, a contour and a label actually mean, and what 2D distance does not.
Work the map
Explore
Move through the space: the map, the cluster tree, categories, bags, journeys and gradients.
Inspect
Read what is behind a selection: the Inspector, Reader, Summarizer, keywords and vignettes.
Analyze
Plots, correlation matrices, point similarity, and notebooks wired into the workspace.
Tools
The map tool island: Select, Lasso, Magic Lasso, Path, Landmark, Semantic Ruler and more.
Agents
The Composer: modes, runtimes, slash commands, custom agents and the shared workspace.
Automate
Macros, projections, selector training, and the Jarvis workflow builder.
Build on Mantis
npm: mantisai-cli
Mantis CLI
Drive your spaces from the terminal, and give Claude Code, Codex or OpenCode direct access to them.
Beta
Extensions
Ship custom panels and verticals with a manifest, a sandboxed UI and an optional Python backend.
Reference
Glossary, keyboard shortcuts, the full panel index, and the repositories behind Mantis.
New to the vocabulary? A space holds your data and one or more maps. A map lays points out so that proximity means similarity. A bag is a saved set of points. Everything else builds on those four words. The glossary has the rest.