Last modified June 21, 2026

Introduction to Muster

Muster is a universal control plane built on the Model Context Protocol (MCP). It aggregates multiple MCP servers behind a single interface and gives AI agents intelligent tool discovery, OAuth-based authentication, workflow orchestration, and dynamic tool loading.

The MCP server sprawl problem

MCP lets an AI assistant call tools exposed by an MCP server—for example, “list the pods in this namespace” against a Kubernetes cluster. That works well for a single server. It breaks down quickly once you have many.

A Giant Swarm customer typically operates a fleet: several management clusters, each with its own Kubernetes API, plus supporting systems such as Prometheus for metrics or Teleport for access. Wiring an assistant directly to these means:

• A separate MCP server connection to configure and maintain for every cluster and every system.
• Separate credentials and authentication flows for each one.
• Every server’s full tool list loaded into the assistant’s context window at once—hundreds of tools, most of them irrelevant to the task at hand. That pollutes the context and drives up token cost on every interaction.

Intelligent aggregation

Muster acts as a meta-MCP server: a single aggregation point that manages many downstream MCP servers and presents their combined capabilities through one connection. Your AI assistant connects to Muster, not to each server individually.

Three properties make the aggregation “intelligent” rather than a simple proxy:

• Conflict-free naming. When two servers expose a tool with the same name, Muster prefixes external tools with their server name, so a tool such as x_kubernetes_get_pods or x_prometheus_query never collides.
• A small, stable tool surface. Instead of exposing every underlying tool, Muster exposes a handful of meta-tools. Agents discover and invoke the full set of capabilities on demand, so the context stays lean and adding or removing a downstream server doesn’t change the assistant’s configuration.
• Live lifecycle management. Muster manages downstream server processes, monitors their health, and updates its tool registry dynamically—no IDE restart required when capabilities change.

What you get

• One endpoint, one login. Authenticate once via your enterprise SSO and reach your whole fleet. Muster handles forwarding and exchanging tokens to downstream servers and clusters transparently—see Security.
• Lower cost. Because agents only load the tools they actually use, and because multi-step operations can be packaged as single workflow calls, the token cost of an AI interaction drops substantially.
• Capabilities beyond Kubernetes. Any MCP server can sit behind Muster—metrics, dashboards, access brokers, and custom internal tools—all reachable through the same connection.

Continue with the architecture to see how the gateway is structured, or jump straight to setting up your AI agent.