Overview

This MCP (Model Context Protocol) server helps you discover and manage local model-serving endpoints on your machine. It exposes lightweight metadata about available model servers so MCP-aware clients can automatically find, register, and use local instances without manual configuration. This is useful for local development, testing, or running multiple model servers concurrently on the same host.

The server is intentionally minimal: it publishes a simple manifest for each MCP instance and provides simple discovery mechanisms so tools and IDEs can enumerate available model endpoints. You can run it as a single binary, container, or as a background service and configure what it advertises via a small JSON/YAML manifest.

Features

Automatic local discovery of MCP server instances for MCP-aware clients
Simple, human-readable manifest format (JSON/YAML) describing endpoints and capabilities
Optional TLS support and custom port configuration
Runs as a standalone binary, Docker container, or background service
Lightweight and suitable for development, CI, and multi-model management
API endpoints for listing advertised instances and fetching metadata

Installation / Configuration

The repository provides multiple ways to run the server. The exact commands depend on the release artifacts in the repo; below are generic examples you can adapt.

Clone the repository:

git clone https://github.com/particlefuture/1mcpserver.git
cd 1mcpserver

Run the included binary (if present):

# If the repository has a prebuilt executable named mcp-server
./mcp-server --config ./examples/local-manifest.yaml

Run with Docker (replace image/name with the repo image if published):

docker run -p 8080:8080 \
  -v $(pwd)/examples/local-manifest.yaml:/etc/mcp/manifest.yaml:ro \
  ghcr.io/particlefuture/1mcpserver:latest \
  --config /etc/mcp/manifest.yaml

Example manifest (YAML)

server:
  name: "local-mcp"
  host: "127.0.0.1"
  port: 8080
  tls: false
models:
  - id: "gpt-local"
    type: "llm"
    endpoint: "http://127.0.0.1:1234/v1"
    description: "Local LLM for development"
    capabilities:
      - completion
      - chat
  - id: "vision-local"
    type: "vision"
    endpoint: "http://127.0.0.1:2345/predict"
    description: "Local vision server"

Minimal JSON manifest

{
  "server": {
    "name": "local-mcp",
    "host": "127.0.0.1",
    "port": 8080
  },
  "models": [
    {
      "id": "gpt-local",
      "type": "llm",
      "endpoint": "http://127.0.0.1:1234/v1",
      "capabilities": ["completion","chat"]
    }
  ]
}

Common options (example CLI)

./mcp-server --config ./manifest.yaml --discovery mdns --log-level info

Configuration key reference

Key	Type	Description
server.name	string	Human-friendly server name
server.host	string	Host IP or hostname to advertise
server.port	integer	Port the MCP server listens on
server.tls	boolean	Whether to use TLS for the advertised endpoints
models[].id	string	Unique model identifier
models[].type	string	Model category (e.g., llm, vision)
models[].endpoint	string	Full URL to the model server API
models[].capabilities	array	Supported operations (completion, chat, predict)

Systemd unit example

[Unit]
Description=MCP Local Discovery Server
After=network.target

[Service]
Type=simple
ExecStart=/usr/local/bin/mcp-server --config /etc/mcp/manifest.yaml
Restart=on-failure

[Install]
WantedBy=multi-user.target

Available Resources

Repository: https://github.com/particlefuture/1mcpserver
Issues and feature requests: use the repo’s Issues page
Example manifests and sample model entries are provided in the repo’s examples/ directory
API endpoints (typical):
- GET /instances — list advertised MCP instances
- GET /instances/{id} — fetch metadata/manifest for an instance (Confirm exact paths in the repo’s documentation)

If you need specific discovery behavior (mDNS, local socket, or polling), check the repo README

Automated Local MCP Server Discovery and Configuration

Overview

Features

Installation / Configuration

Available Resources