The research layer for AI agents
Give agents a workspace, set permissions, and share findings, all on the same map your team works in.
Agents that read the map, and cite what they used
Step through one Composer turn: the question arrives, the agent plans, runs a semantic search and reads a cluster through Mantis tools, gathers the record ids behind its answer, and hands the answer back with citations the researcher can open on the map.
Orchestrate your agents
Run several agents over the same map with the Mantis plugin system. Plugins are the open interface for custom tools, letting agents search, filter, and summarize points, build bags, or call your own services. Every result lands in the map where your team can inspect it, and every step stays traceable to the points it used.
- Multi-agent runsOrchestrate several agents over one map
- Live suggestionsGet next-action suggestions as the map changes
- Custom toolsAdd your own tools through the plugin system
Give agents a workspace
Use a Mantis space to let your agents open maps, run searches, and complete tasks within scope. Give each agent access to specific maps, bags, and fields, and set limits on what it can read or write. Or let your applications create user, agent, and team accounts on Mantis with the Mantis Accounts SDK.
- Flexible accessLet agents work through long investigations without a human driving every step
- Intuitive controlsSet scope per agent or workflow and control what each one can read
- Low overheadAgents reuse the embeddings your team already has
Learn more
Composer
The AI chat panel that reads your space, calls Mantis tools, and drives the map.
Read more
Agent Studio and memories
Build a custom agent with its own knowledge and MCP tools, and manage what the Composer remembers about you.
Read more
Mantis CLI
Give any coding agent access to your Mantis workspace from the terminal, without an editor MCP plugin.
Read more