Concepts

The core ideas behind Kubemoot: crews and agents, the consensus model, signals, models and scheduling, MCP tools, and conversations, turns, and threads.

The core ideas behind Kubemoot’s consensus model.

  • The Moot - Consensus Model : How a crew settles a question by deliberation, and how that deliberation can be organized.
  • Conversations, Turns & Threads : The units of interaction: a durable conversation, the turns within it, and the thread each turn's discussion runs on.
  • Crews & Agents : Crews, coordinators, Toolers, Analysts, and how they compose.
  • Agent Context : Where an agent's model context comes from in Kubemoot: what each source provides today, how it is written and read, how long it lives, how it is isolated, and which sources are trusted instructions and which are untrusted data.
  • The Realm and the Crew : Kubemoot's resources fall into two layers: the realm a crew draws on, and the crew itself. Which kinds sit where, and why a crew stays portable.
  • The Table and the Harnesses : A crew as a harness made of harnesses: what Kubemoot builds at the table, what it adopts, and what it convenes.
  • ADL, the Architecture Definition Language, for agents : What ADL is, where it comes from, and why Kubemoot writes agent behavior in it.
  • Signals & Protocol : agree / concern / block / stand_aside and the discussion phases.
  • The Workspace : How Kubemoot stores every tool result as its own file in the discussion's workspace, puts the file's description in the discussion, and posts a result as well when it is small enough.
  • Models & Scheduling : Loose model coupling and just-in-time accelerator scheduling.
  • Choosing a Model : How to pick models for a crew when accelerator memory is the binding constraint.
  • MCP Tools : How agents gain capabilities through MCP servers and the gateway.
  • Secrets and tools : Why the model never holds a secret, how credentials reach tools, and how the use grant and its admission check decide who may reference which Secret.
  • Network policies : How the operator generates NetworkPolicy objects for a crew: default-deny ingress, an allow for each tool server, and outbound rules for the workloads that declare them.
  • Isolation : How to run agent and MCP server pods under a sandboxing runtime such as gVisor with a Kubernetes RuntimeClass, why it is optional, and what the RuntimeClassAvailable condition says.
  • Code sandbox : How a crew gets exact counts, sums, and sorts by running short programs in the code sandbox, how each run is contained, and how to run its pod under a sandboxed runtime such as gVisor.
  • Crew threat model : How injected instructions travel between agents in a crew, which defenses Kubemoot applies today, what no defense covers yet, and the fitness tests that measure how a crew holds up.

The Moot - Consensus Model

How a crew settles a question by deliberation, and how that deliberation can be organized.

Conversations, Turns & Threads

The units of interaction: a durable conversation, the turns within it, and the thread each turn’s discussion runs on.

Crews & Agents

Crews, coordinators, Toolers, Analysts, and how they compose.

Agent Context

Where an agent’s model context comes from in Kubemoot: what each source provides today, how it is written and read, how long it lives, how it is isolated, and which sources are trusted instructions and which are untrusted data.

The Realm and the Crew

Kubemoot’s resources fall into two layers: the realm a crew draws on, and the crew itself. Which kinds sit where, and why a crew stays portable.

The Table and the Harnesses

A crew as a harness made of harnesses: what Kubemoot builds at the table, what it adopts, and what it convenes.

ADL, the Architecture Definition Language, for agents

What ADL is, where it comes from, and why Kubemoot writes agent behavior in it.

Signals & Protocol

agree / concern / block / stand_aside and the discussion phases.

The Workspace

How Kubemoot stores every tool result as its own file in the discussion’s workspace, puts the file’s description in the discussion, and posts a result as well when it is small enough.

Models & Scheduling

Loose model coupling and just-in-time accelerator scheduling.

Choosing a Model

How to pick models for a crew when accelerator memory is the binding constraint.

MCP Tools

How agents gain capabilities through MCP servers and the gateway.

Secrets and tools

Why the model never holds a secret, how credentials reach tools, and how the use grant and its admission check decide who may reference which Secret.

Network policies

How the operator generates NetworkPolicy objects for a crew: default-deny ingress, an allow for each tool server, and outbound rules for the workloads that declare them.

Isolation

How to run agent and MCP server pods under a sandboxing runtime such as gVisor with a Kubernetes RuntimeClass, why it is optional, and what the RuntimeClassAvailable condition says.

Code sandbox

How a crew gets exact counts, sums, and sorts by running short programs in the code sandbox, how each run is contained, and how to run its pod under a sandboxed runtime such as gVisor.

Crew threat model

How injected instructions travel between agents in a crew, which defenses Kubemoot applies today, what no defense covers yet, and the fitness tests that measure how a crew holds up.