Gridctl is an MCP orchestration tool and server tool designed for AI agent development. It acts as a gateway, connecting to and orchestrating various Model Context Protocol servers that interact with protocols like OpenADR and OCPP. Gridctl normalizes diverse communication methods from these servers, exposing them through a single endpoint for Large Language Model clients. It supports multiple transport protocols, including stdio, HTTP, SSE, and servers behind SSH tunnels, unifying access for compatible AI clients like Claude Desktop and Cursor. This tool optimizes the developer loop for local AI agent experimentation.

What does Gridctl connect to?

Gridctl connects to and orchestrates various Model Context Protocol (MCP) servers. These MCP servers can, in turn, interact with external protocols such as OpenADR and OCPP. Gridctl functions as a central gateway, bridging Large Language Model (LLM) clients with these diverse MCP servers. It normalizes their varied communication methods, exposing them all through a single, unified endpoint, typically localhost:8180. This architecture simplifies how LLM clients access tools from multiple backend services.

Core features and tools

Gridctl provides a command-line interface (CLI) with specific callable actions. These include gridctl link <client_name> for connecting AI clients like "claude" or "cursor", gridctl deploy stack.yaml for deploying an MCP server stack, and gridctl reload for hot reloading configurations. While Gridctl orchestrates and aggregates tools from other MCP servers for AI clients, it does not expose its own functions directly as Model Context Protocol (MCP) tools callable by an LLMs.

Who is it for?

Gridctl is primarily for developers and operators engaged in AI agent development and iteration. It is specifically designed to optimize the "developer loop" and enhance local development environments. The tool supports rapid experimentation and management of AI agent infrastructure. It simplifies the process of configuring and connecting various Model Context Protocol servers for AI client interaction. Skip this if your primary need is for a tool that exposes its own LLM-callable functions, rather than orchestrating external ones.

How to get started

Authentication for Gridctl involves managing secrets in an encrypted store, addressing the issue of API keys and tokens in plaintext client configuration files. The setup defines the entire AI tool infrastructure, including MCP servers and their configurations, in a single stack.yaml file.

Use cases

• Rapidly deploying, testing, and tearing down ephemeral MCP server stacks for AI agent experimentation.
• Consolidating and orchestrating multiple MCP servers by defining them in a single declarative stack.yaml file, simplifying infrastructure management.
• Providing a unified gateway (e.g., localhost:8180) for various LLM clients to access a curated set of tools from multiple MCP servers without complex individual client configurations.

Gridctl offers a focused solution for orchestrating Model Context Protocol servers and connecting them to Large Language Model clients. It addresses secure credential management and simplifies infrastructure definition through a declarative stack.yaml file. Tailored for local AI agent development, it unifies access to diverse server-side tools. Try connecting a compatible AI client like Claude Desktop using "gridctl link claude" to begin orchestrating your MCP servers.