NAB01–05
Five container-based computer environments with internal MCP tooling. Each Neuro-Agentic Base gives an LLM and its harness a workspace to build in; bases can connect to form Hive-Minds.
Project illustrationThe challenge
Give an agent a coherent computer environment for building and running work, then connect multiple environments so tasks and results can move between them.
The approach
Package the workspace and its tool interfaces inside each NAB container, including terminal, browser, files and development tools.
Let the language model reason through a harness that invokes the internal MCP tools and returns observations from the environment.
Connect NAB01–05 through task and result hand-offs to form a Hive-Mind, while each base retains its own computer workspace.
Use the interactive architecture picture to inspect the shared layers and switch between independent and connected operation. It illustrates the design rather than a live fleet status.
The architecture
- A consistent architecture from LLM reasoning through the harness and MCP tools to a container-based computer workspace.
- A shared design for five Neuro-Agentic Bases, with independent work and coordinated Hive-Mind operation.
- A container provides the VM-like workspace experience described here; deployment-specific permissions and resource limits determine its operating boundary.
