M.o.E / Neuro-Agentic Base
Five bases.
Connected intelligence.
NAB01–05 give agents a computer environment to work in. An LLM reasons, a harness executes, and internal MCP tooling opens the workspace. Connect the bases to form a Hive-Mind.
NAB / Neuro-Agentic Base
An agent with
a computer to build in.
A computer environment inside a container: tools, files and a workspace the agent can operate.
One base can work alone.
Together, they can form a Hive-Mind.
Connect the NABs to coordinate tasks, exchange results and combine specialist work. Each base retains its own computer workspace and internal tools.
THE COMPUTER INSIDE THE CONTAINER
More than a chat window.
A Neuro-Agentic Base brings the working environment around an agent together: terminal, filesystem, browser, desktop and development tools. Through its harness and internal MCP interfaces, the model can inspect, build, run and refine work inside that environment.
The experience is VM-like: a full computer workspace available to the agent. The architecture uses containers; the diagram does not imply a separate virtual machine or hardware isolation for every base.
HIVE-MIND CONNECTIONS
Separate workspaces.
Shared objectives.
Connected NABs coordinate work through task hand-offs, exchanged results and tool-mediated communication. Each base keeps its own environment while contributing to the larger task.
The selector illustrates the common architecture of NAB01–05. Instance roles, permissions and active connections depend on the configured deployment.
Related foundations
Control has a context.
Sandy’s harness operates a physical Linux host. Hopper uses bounded inspection jobs with human review. Jinni explores personal context and approved memory. Each case study explains the scope of its own environment and tools.
Personal agent commissions
A system for your work.
Discuss the computer environment, tools and operating boundaries your agent would need.
Discuss an agentThe dedicated agent business is being developed at lwc.moe.