
mcp_sse
An elixir Model Context Protocal (MCP) server library which uses the Server-Sent Events (SSE) transport type
3 years
Works with Finder
2
Github Watches
10
Github Forks
41
Github Stars
MCP over SSE
This library provides a simple implementation of the Model Context Protocol (MCP) over Server-Sent Events (SSE).
For more information about the Model Context Protocol, visit: Model Context Protocol Documentation
Table of Contents
- Features
- Create your own MCP server
- Installation
- Usage
- Configuration Options
- Quick Demo
- Other Notes
- Contributing
Features
- Full MCP server implementation
- SSE connection management
- JSON-RPC message handling
- Tool registration and execution
- Session management
- Automatic ping/keepalive
- Error handling and validation
Create your own MCP server
You must implement the MCPServer
behaviour.
You only need to implement the required callbacks (handle_ping/1
and handle_initialize/2
) and any optional callbacks for features you want to support.
The use MCPServer
macro provides:
- Built-in message routing
- Protocol version validation
- Default implementations for optional callbacks
- JSON-RPC error handling
- Logging
See DefaultServer
for a default implementation of the MCPServer
behaviour.
Installation
For Phoenix Applications:
- Add the required configuration to
config/config.exs
:
# Configure MIME types for SSE
config :mime, :types, %{
"text/event-stream" => ["sse"]
}
# Configure the MCP Server
config :mcp_sse, :mcp_server, YourApp.YourMCPServer
- Add to your dependencies in
mix.exs
:
def deps do
[
{:mcp_sse, "~> 0.1.4"}
]
end
- Configure your router (
lib/your_app_web/router.ex
):
pipeline :sse do
plug :accepts, ["sse"]
end
scope "/" do
pipe_through :sse
get "/sse", SSE.ConnectionPlug, :call
post "/message", SSE.ConnectionPlug, :call
end
- Run your application:
mix phx.server
For Plug Applications with Bandit:
- Create a new Plug application with supervision:
mix new your_app --sup
- Add the required configuration to
config/config.exs
:
import Config
# Configure MIME types for SSE
config :mime, :types, %{
"text/event-stream" => ["sse"]
}
# Configure the MCP Server
config :mcp_sse, :mcp_server, YourApp.YourMCPServer
- Add dependencies to
mix.exs
:
def deps do
[
{:mcp_sse, "~> 0.1.4"},
{:plug, "~> 1.14"},
{:bandit, "~> 1.2"}
]
end
- Configure your router (
lib/your_app/router.ex
):
defmodule YourApp.Router do
use Plug.Router
plug Plug.Parsers,
parsers: [:urlencoded, :json],
pass: ["text/*"],
json_decoder: JSON
plug :match
plug :ensure_session_id
plug :dispatch
# Middleware to ensure session ID exists
def ensure_session_id(conn, _opts) do
case get_session_id(conn) do
nil ->
# Generate a new session ID if none exists
session_id = generate_session_id()
%{conn | query_params: Map.put(conn.query_params, "sessionId", session_id)}
_session_id ->
conn
end
end
# Helper to get session ID from query params
defp get_session_id(conn) do
conn.query_params["sessionId"]
end
# Generate a unique session ID
defp generate_session_id do
Base.encode16(:crypto.strong_rand_bytes(8), case: :lower)
end
forward "/sse", to: SSE.ConnectionPlug
forward "/message", to: SSE.ConnectionPlug
match _ do
send_resp(conn, 404, "Not found")
end
end
- Set up your application supervision (
lib/your_app/application.ex
):
defmodule YourApp.Application do
use Application
@impl true
def start(_type, _args) do
children = [
{Bandit, plug: YourApp.Router, port: 4000}
]
opts = [strategy: :one_for_one, name: YourApp.Supervisor]
Supervisor.start_link(children, opts)
end
end
- Run your application:
mix run --no-halt
Usage
With MCP Inspector
- Start the inspector:
MCP_SERVER_URL=localhost:4000 npx @modelcontextprotocol/inspector@latest
- Navigate to http://localhost:6274/
- Make sure your server is running
- Click
Connect
- You can now list tools and call them
With Cursor
- Open Cursor Settings
- Navigate to the MCP tab
- Click
Add new global MCP server
- Fill in the
~/.cursor/mcp.json
with:
{
"mcpServers": {
"your-mcp-server": {
"url": "http://localhost:4000/sse"
}
}
}
- Make sure your server is running
- Ask Cursor to run one of your tools
Configuration Options
The Bandit server can be configured with additional options in your application module:
# Example with custom port and HTTPS
children = [
{Bandit,
plug: YourApp.Router,
port: System.get_env("PORT", "4000") |> String.to_integer(),
scheme: :https,
certfile: "priv/cert/selfsigned.pem",
keyfile: "priv/cert/selfsigned_key.pem"
}
]
Quick Demo
To see the MCP server in action:
- Start a server in one terminal:
# Our example server
elixir dev/example_server.exs
# Your Phoenix application
mix phx.server
# Your Plug application
mix run --no-halt
- In another terminal, run the demo client script:
elixir dev/example_client.exs
The client script will:
- Connect to the SSE endpoint
- Initialize the connection
- List available tools
- Call the upcase tool with example input
- Display the results of each step
This provides a practical demonstration of the Model Context Protocol flow and server capabilities.
Other Notes
Example Client Usage
// Connect to SSE endpoint
const sse = new EventSource('/sse');
// Handle endpoint message
sse.addEventListener('endpoint', (e) => {
const messageEndpoint = e.data;
// Use messageEndpoint for subsequent JSON-RPC requests
});
// Send initialize request
fetch('/message?sessionId=YOUR_SESSION_ID', {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify({
jsonrpc: '2.0',
id: 1,
method: 'initialize',
params: {
protocolVersion: '2024-11-05',
capabilities: {}
}
})
});
Session Management
The MCP SSE server requires a session ID for each connection. The router automatically:
- Uses an existing session ID from query parameters if provided
- Generates a new session ID if none exists
- Ensures all requests to
/sse
and/message
endpoints have a valid session ID
SSE Keepalive
The SSE connection sends periodic keepalive pings to prevent connection timeouts.
You can configure the ping interval or disable it entirely in config/config.exs
:
# Set custom ping interval (in milliseconds)
config :mcp_sse, :sse_keepalive_timeout, 30_000 # 30 seconds
# Or disable pings entirely
config :mcp_sse, :sse_keepalive_timeout, :infinity
MCP Response Formatting
When implementing tool responses in your MCP server, the content must follow the MCP specification for content types. The response content should be formatted as one of these types:
# Text content
{:ok,
%{
jsonrpc: "2.0",
id: request_id,
result: %{
content: [
%{
type: "text",
text: "Your text response here"
}
]
}
}}
# Image content
{:ok,
%{
jsonrpc: "2.0",
id: request_id,
result: %{
content: [
%{
type: "image",
data: "base64_encoded_image_data",
mimeType: "image/png"
}
]
}
}}
# Resource reference
{:ok,
%{
jsonrpc: "2.0",
id: request_id,
result: %{
content: [
%{
type: "resource",
resource: %{
name: "resource_name",
description: "resource description"
}
}
]
}
}}
For structured data like JSON, you should convert it to a formatted string:
def handle_call_tool(request_id, %{"name" => "list_companies"} = _params) do
companies = fetch_companies() # Your data fetching logic
{:ok,
%{
jsonrpc: "2.0",
id: request_id,
result: %{
content: [
%{
type: "text",
text: JSON.encode!(companies, pretty: true)
}
]
}
}}
end
For more details on response formatting, see the MCP Content Types Specification.
Contributing
- Fork the repository and clone it
- Create a new branch in your fork
- Make your changes and commit them
- Push the changes to your fork
- Open a pull request in upstream
相关推荐
I find academic articles and books for research and literature reviews.
Converts Figma frames into front-end code for various mobile frameworks.
Confidential guide on numerology and astrology, based of GG33 Public information
Embark on a thrilling diplomatic quest across a galaxy on the brink of war. Navigate complex politics and alien cultures to forge peace and avert catastrophe in this immersive interstellar adventure.
Advanced software engineer GPT that excels through nailing the basics.
Delivers concise Python code and interprets non-English comments
💬 MaxKB is a ready-to-use AI chatbot that integrates Retrieval-Augmented Generation (RAG) pipelines, supports robust workflows, and provides advanced MCP tool-use capabilities.
Micropython I2C-based manipulation of the MCP series GPIO expander, derived from Adafruit_MCP230xx
The all-in-one Desktop & Docker AI application with built-in RAG, AI agents, No-code agent builder, MCP compatibility, and more.
MCP server to provide Figma layout information to AI coding agents like Cursor
AI Agents & MCPs & AI Workflow Automation • (280+ MCP servers for AI agents) • AI Automation / AI Agent with MCPs • AI Workflows & AI Agents • MCPs for AI Agents
Python code to use the MCP3008 analog to digital converter with a Raspberry Pi or BeagleBone black.
Reviews

user_nzpExbJQ
I've been using the mcp_sse by kEND, and it's been a game-changer for me. The seamless integration and user-friendly design make it a must-have for anyone in need of streamlined server-sent events. It's evident that a lot of thought went into its development. Highly recommend checking it out! Link: https://github.com/kEND/mcp_sse