# tigranbs/mcgravity

**Category:** 🔗 Aggregators  
**Repository:** https://github.com/tigranbs/mcgravity  
**GitHub Stars:** 102  
**Views:** 3  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/tigranbs-mcgravity

## Description
A proxy tool for composing multiple MCP servers into one unified endpoint. Scale your AI tools by load balancing requests across multiple MCP servers, similar to how Nginx works for web servers.

## Claude Desktop Quick Installation
Heuristic fallback — verify the package name and runner against the repository README before running it. Uses `npx` (confidence: low):

```json
"mcpServers": {
  "mcgravity": {
    "command": "npx",
    "args": ["-y","tigranbs-mcgravity"]
  }
}
```

## Documentation

## What tigranbs/mcgravity MCP server does

The tigranbs/mcgravity MCP server is a proxy for composing multiple MCP servers into one unified endpoint. Instead of exposing each server independently to an AI client, it places a single layer in front of the group. That layer represents the collection of MCP servers as one destination.

Its primary use is request distribution. Multiple MCP servers can be placed behind the proxy, allowing incoming requests to be load balanced across them. The repository description compares this operating model to Nginx distributing traffic across web servers. This makes the project relevant to deployments where one MCP endpoint must front multiple server processes or instances.

## How it works

The proxy receives requests intended for the composed MCP setup and routes them across the configured MCP servers. The available material identifies load balancing as the central routing behavior, but does not specify the balancing algorithm, health checks, retry behavior, session handling, transport details, or failure handling.

The tigranbs/mcgravity MCP server is therefore best understood as an aggregation and distribution layer, not as an MCP server that adds a particular business-domain tool. Its value depends on the MCP servers connected behind it and on the client’s ability to use the resulting unified endpoint.

No README or configuration reference is available in the supplied material. As a result, the exact topology, routing rules, supported transports, and operational requirements cannot be documented here without risking unsupported claims.

## Setup and configuration

The supplied description does not include installation commands, package names, container images, configuration files, command-line options, or environment variables. Installation details should be taken from the repository’s own documentation before deployment.

The description also does not state whether the proxy requires credentials, how backend MCP servers are registered, or whether the unified endpoint is local or remotely hosted. Those details matter when connecting an AI client, but they cannot be determined from the available information.

## Tools and capabilities

The tigranbs/mcgravity MCP server provides the following documented capability:

- Composes multiple MCP servers behind one endpoint.
- Proxies requests between the unified endpoint and backend MCP servers.
- Load balances requests across multiple MCP servers.
- Supports scaling an AI tool deployment by distributing traffic across server instances.

The source material does not identify individual MCP tools exposed by the proxy. It describes infrastructure for serving existing MCP tools rather than a named set of domain-specific operations.

## Limitations and notes

The available information does not confirm support for any specific MCP client, transport, authentication method, persistence model, observability feature, or deployment platform. It also does not provide performance characteristics or explain how the proxy behaves when a backend server becomes unavailable.

Use the tigranbs/mcgravity MCP server when the main requirement is aggregation and request distribution across MCP servers. For a production deployment, verify the repository’s current documentation for backend registration, balancing behavior, endpoint configuration, and failure handling before relying on it for capacity or availability.

_Full upstream README: https://allmcps.com/mcp/tigranbs-mcgravity/readme_