agent lab

Manufacture your first agent. Then duplicate the team.

Create an Einstein-style agent with model, role, tools, knowledge, terminal, voice, memory, and a clickable position inside a Platonic team.

agent.tui

connected

        /////////\\
     ////  .--.  \\
   ///   ( o  o )  \\
  |||      /--\      ||
  |||   .-\__/-.     ||
   \\    \____/     //
    \\  /|    |\   //
      \/_|____|_\_//
        /  ||  \
       /___||___\

terminal

@einstein:~$ run mission --trace

status

modelGPT-4.1

rolePlanner

terminalenabled

voicecapturable

modeleader/router

tools

terminalsearchfiles

context

> foundry.md

> workflow.txt

team topology

Pentachoron entry cell

click vertex to chat

edges are handoffs

white node is selected

ring node is router

agent editor

agent chat

Einstein

operator prefers concise traces

voice capture

Hold to record. Release to analyze and attach to memory.

knowledge ingestion

foundry.mdembeddedlinked

workflow.txtembeddedlinked

tools and integrations

OpenAI

LLM provider. Neutral label until official logo asset is approved.

Anthropic

LLM provider. Neutral label until official logo asset is approved.

Gemini

LLM provider. Neutral label until official logo asset is approved.

message channel. Neutral label until official logo asset is approved.

voice/chat channel. Neutral label until official logo asset is approved.

Google Drive

knowledge source. Neutral label until official logo asset is approved.

Local CLI

tool/runtime. Neutral label until official logo asset is approved.

cli export

$ numenbox install team --topology pentachoron

$ numenbox run team --workflow "turn inbound leads into qualified meetings"

$ numenbox inspect trace --agent einstein

$ numenbox deploy --channel cli --channel whatsapp

price Pentachoron->

agent interaction viewer

Watch the lab work happen.

trace timeline

Einstein -> Researcher 02

search / foundry.md

Researcher 02 -> Critic 03

files / workflow.txt

Critic 03 -> Einstein

terminal / trace.log

message log

Einstein requests market facts for turn inbound leads into qualified meetings.

Researcher 02 sends enriched context and asks for critique.

Critic 03 returns verified artifact for final assembly.

tool log

$ search --agent Einstein

$ files --agent Researcher 02

$ terminal --agent Critic 03

final artifact

Einstein assembled the team output: qualified workflow, tool calls, evidence, and deploy command.