Add everything-server for MCP conformance testing

Implements a comprehensive reference server that exercises all MCP protocol
features for conformance testing. This server validates that the Python SDK
correctly implements the MCP specification by passing all 25 conformance
test scenarios.

Features implemented:
- 11 tools covering all content types (text, image, audio, embedded resources,
  mixed content, logging, progress, error handling, sampling, elicitation)
- 4 resources (static text/binary, templates, subscribable resources)
- 4 prompts with various content types
- Logging support (logging/setLevel)
- Completion support (completion/complete)
- Resource subscriptions (resources/subscribe, resources/unsubscribe)
- Session management with StreamableHTTP transport

Implementation uses FastMCP with proper type safety:
- All URIs wrapped with AnyUrl() for type correctness
- Context properly typed with ServerSession parameter
- Subscribe/unsubscribe handlers use correct AnyUrl parameter and None return
- Completion handler returns Completion objects
- ElicitationResult uses discriminated union narrowing via action field
- Relative imports in __main__.py to avoid type stub warnings

Type checking: Passes pyright with 0 errors (only 3 targeted ignores for
accessing _mcp_server private API, documented with TODO to add public APIs
to FastMCP in the future).

Conformance: Passes all 25 test scenarios.

Author: felixweinberger

examples/servers/everything-server/README.md

@@ -0,0 +1,42 @@
+# MCP Everything Server
+
+A comprehensive MCP server implementing all protocol features for conformance testing.
+
+## Overview
+
+The Everything Server is a reference implementation that demonstrates all features of the Model Context Protocol (MCP). It is designed to be used with the [MCP Conformance Test Framework](https://github.com/modelcontextprotocol/conformance) to validate MCP client and server implementations.
+
+## Installation
+
+From the python-sdk root directory:
+
+```bash
+uv sync --frozen
+```
+
+## Usage
+
+### Running the Server
+
+Start the server with default settings (port 3001):
+
+```bash
+uv run -m mcp_everything_server
+```
+
+Or with custom options:
+
+```bash
+uv run -m mcp_everything_server --port 3001 --log-level DEBUG
+```
+
+The server will be available at: `http://localhost:3001/mcp`
+
+### Command-Line Options
+
+- `--port` - Port to listen on (default: 3001)
+- `--log-level` - Logging level: DEBUG, INFO, WARNING, ERROR, CRITICAL (default: INFO)
+
+## Running Conformance Tests
+
+See the [MCP Conformance Test Framework](https://github.com/modelcontextprotocol/conformance) for instructions on running conformance tests against this server.

examples/servers/everything-server/mcp_everything_server/__init__.py

@@ -0,0 +1,3 @@
+"""MCP Everything Server - Comprehensive conformance test server."""
+
+__version__ = "0.1.0"

examples/servers/everything-server/mcp_everything_server/__main__.py

@@ -0,0 +1,6 @@
+"""CLI entry point for the MCP Everything Server."""
+
+from .server import main
+
+if __name__ == "__main__":
+    main()

examples/servers/everything-server/mcp_everything_server/server.py

@@ -0,0 +1,319 @@
+#!/usr/bin/env python3
+"""
+MCP Everything Server - Conformance Test Server
+
+Server implementing all MCP features for conformance testing based on Conformance Server Specification.
+"""
+
+import asyncio
+import logging
+
+import click
+from mcp.server.fastmcp import Context, FastMCP
+from mcp.server.fastmcp.prompts.base import UserMessage
+from mcp.server.session import ServerSession
+from mcp.types import (
+    AudioContent,
+    Completion,
+    CompletionArgument,
+    CompletionContext,
+    EmbeddedResource,
+    ImageContent,
+    PromptReference,
+    ResourceTemplateReference,
+    SamplingMessage,
+    TextContent,
+    TextResourceContents,
+)
+from pydantic import AnyUrl, BaseModel, Field
+
+logger = logging.getLogger(__name__)
+
+# Test data
+TEST_IMAGE_BASE64 = "iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVR42mP8z8DwHwAFBQIAX8jx0gAAAABJRU5ErkJggg=="
+TEST_AUDIO_BASE64 = "UklGRiYAAABXQVZFZm10IBAAAAABAAEAQB8AAAB9AAACABAAZGF0YQIAAAA="
+
+# Server state
+resource_subscriptions: set[str] = set()
+watched_resource_content = "Watched resource content"
+
+mcp = FastMCP(
+    name="mcp-conformance-test-server",
+)
+
+
+# Tools
+@mcp.tool()
+def test_simple_text() -> str:
+    """Tests simple text content response"""
+    return "This is a simple text response for testing."
+
+
+@mcp.tool()
+def test_image_content() -> list[ImageContent]:
+    """Tests image content response"""
+    return [ImageContent(type="image", data=TEST_IMAGE_BASE64, mimeType="image/png")]
+
+
+@mcp.tool()
+def test_audio_content() -> list[AudioContent]:
+    """Tests audio content response"""
+    return [AudioContent(type="audio", data=TEST_AUDIO_BASE64, mimeType="audio/wav")]
+
+
+@mcp.tool()
+def test_embedded_resource() -> list[EmbeddedResource]:
+    """Tests embedded resource content response"""
+    return [
+        EmbeddedResource(
+            type="resource",
+            resource=TextResourceContents(
+                uri=AnyUrl("test://embedded-resource"),
+                mimeType="text/plain",
+                text="This is an embedded resource content.",
+            ),
+        )
+    ]
+
+
+@mcp.tool()
+def test_multiple_content_types() -> list[TextContent | ImageContent | EmbeddedResource]:
+    """Tests response with multiple content types (text, image, resource)"""
+    return [
+        TextContent(type="text", text="Multiple content types test:"),
+        ImageContent(type="image", data=TEST_IMAGE_BASE64, mimeType="image/png"),
+        EmbeddedResource(
+            type="resource",
+            resource=TextResourceContents(
+                uri=AnyUrl("test://mixed-content-resource"),
+                mimeType="application/json",
+                text='{"test": "data", "value": 123}',
+            ),
+        ),
+    ]
+
+
+@mcp.tool()
+async def test_tool_with_logging(ctx: Context[ServerSession, None]) -> str:
+    """Tests tool that emits log messages during execution"""
+    await ctx.info("Tool execution started")
+    await asyncio.sleep(0.05)
+
+    await ctx.info("Tool processing data")
+    await asyncio.sleep(0.05)
+
+    await ctx.info("Tool execution completed")
+    return "Tool with logging executed successfully"
+
+
+@mcp.tool()
+async def test_tool_with_progress(ctx: Context[ServerSession, None]) -> str:
+    """Tests tool that reports progress notifications"""
+    await ctx.report_progress(progress=0, total=100, message="Completed step 0 of 100")
+    await asyncio.sleep(0.05)
+
+    await ctx.report_progress(progress=50, total=100, message="Completed step 50 of 100")
+    await asyncio.sleep(0.05)
+
+    await ctx.report_progress(progress=100, total=100, message="Completed step 100 of 100")
+
+    # Return progress token as string
+    progress_token = ctx.request_context.meta.progressToken if ctx.request_context and ctx.request_context.meta else 0
+    return str(progress_token)
+
+
+class SamplingInput(BaseModel):
+    prompt: str = Field(description="The prompt to send to the LLM")
+
+
+@mcp.tool()
+async def test_sampling(prompt: str, ctx: Context[ServerSession, None]) -> str:
+    """Tests server-initiated sampling (LLM completion request)"""
+    try:
+        # Request sampling from client
+        result = await ctx.session.create_message(
+            messages=[SamplingMessage(role="user", content=TextContent(type="text", text=prompt))],
+            max_tokens=100,
+        )
+
+        if result.content.type == "text":
+            model_response = result.content.text
+        else:
+            model_response = "No response"
+
+        return f"LLM response: {model_response}"
+    except Exception as e:
+        return f"Sampling not supported or error: {str(e)}"
+
+
+class ElicitationInput(BaseModel):
+    message: str = Field(description="The message to show the user")
+
+
+class UserResponse(BaseModel):
+    response: str = Field(description="User's response")
+
+
+@mcp.tool()
+async def test_elicitation(message: str, ctx: Context[ServerSession, None]) -> str:
+    """Tests server-initiated elicitation (user input request)"""
+    try:
+        # Request user input from client
+        result = await ctx.elicit(message=message, schema=UserResponse)
+
+        # Type-safe discriminated union narrowing using action field
+        if result.action == "accept":
+            content = result.data.model_dump_json()
+        else:  # decline or cancel
+            content = "{}"
+
+        return f"User response: action={result.action}, content={content}"
+    except Exception as e:
+        return f"Elicitation not supported or error: {str(e)}"
+
+
+@mcp.tool()
+def test_error_handling() -> str:
+    """Tests error response handling"""
+    raise RuntimeError("This tool intentionally returns an error for testing")
+
+
+# Resources
+@mcp.resource("test://static-text")
+def static_text_resource() -> str:
+    """A static text resource for testing"""
+    return "This is the content of the static text resource."
+
+
+@mcp.resource("test://static-binary")
+def static_binary_resource() -> bytes:
+    """A static binary resource (image) for testing"""
+    import base64
+
+    return base64.b64decode(TEST_IMAGE_BASE64)
+
+
+@mcp.resource("test://template/{id}/data")
+def template_resource(id: str) -> str:
+    """A resource template with parameter substitution"""
+    import json
+
+    return json.dumps({"id": id, "templateTest": True, "data": f"Data for ID: {id}"})
+
+
+@mcp.resource("test://watched-resource")
+def watched_resource() -> str:
+    """A resource that can be subscribed to for updates"""
+    return watched_resource_content
+
+
+# Prompts
+@mcp.prompt()
+def test_simple_prompt() -> list[UserMessage]:
+    """A simple prompt without arguments"""
+    return [UserMessage(role="user", content=TextContent(type="text", text="This is a simple prompt for testing."))]
+
+
+@mcp.prompt()
+def test_prompt_with_arguments(arg1: str, arg2: str) -> list[UserMessage]:
+    """A prompt with required arguments"""
+    return [
+        UserMessage(
+            role="user", content=TextContent(type="text", text=f"Prompt with arguments: arg1='{arg1}', arg2='{arg2}'")
+        )
+    ]
+
+
+@mcp.prompt()
+def test_prompt_with_embedded_resource(resourceUri: str) -> list[UserMessage]:
+    """A prompt that includes an embedded resource"""
+    return [
+        UserMessage(
+            role="user",
+            content=EmbeddedResource(
+                type="resource",
+                resource=TextResourceContents(
+                    uri=AnyUrl(resourceUri),
+                    mimeType="text/plain",
+                    text="Embedded resource content for testing.",
+                ),
+            ),
+        ),
+        UserMessage(role="user", content=TextContent(type="text", text="Please process the embedded resource above.")),
+    ]
+
+
+@mcp.prompt()
+def test_prompt_with_image() -> list[UserMessage]:
+    """A prompt that includes image content"""
+    return [
+        UserMessage(role="user", content=ImageContent(type="image", data=TEST_IMAGE_BASE64, mimeType="image/png")),
+        UserMessage(role="user", content=TextContent(type="text", text="Please analyze the image above.")),
+    ]
+
+
+# Custom request handlers
+# TODO(felix): Add public APIs to FastMCP for subscribe_resource, unsubscribe_resource,
+# and set_logging_level to avoid accessing protected _mcp_server attribute.
+@mcp._mcp_server.set_logging_level()  # pyright: ignore[reportPrivateUsage]
+async def handle_set_logging_level(level: str) -> None:
+    """Handle logging level changes"""
+    logger.info(f"Log level set to: {level}")
+    # In a real implementation, you would adjust the logging level here
+    # For conformance testing, we just acknowledge the request
+
+
+async def handle_subscribe(uri: AnyUrl) -> None:
+    """Handle resource subscription"""
+    resource_subscriptions.add(str(uri))
+    logger.info(f"Subscribed to resource: {uri}")
+
+
+async def handle_unsubscribe(uri: AnyUrl) -> None:
+    """Handle resource unsubscription"""
+    resource_subscriptions.discard(str(uri))
+    logger.info(f"Unsubscribed from resource: {uri}")
+
+
+mcp._mcp_server.subscribe_resource()(handle_subscribe)  # pyright: ignore[reportPrivateUsage]
+mcp._mcp_server.unsubscribe_resource()(handle_unsubscribe)  # pyright: ignore[reportPrivateUsage]
+
+
+@mcp.completion()
+async def _handle_completion(
+    ref: PromptReference | ResourceTemplateReference,
+    argument: CompletionArgument,
+    context: CompletionContext | None,
+) -> Completion:
+    """Handle completion requests"""
+    # Basic completion support - returns empty array for conformance
+    # Real implementations would provide contextual suggestions
+    return Completion(values=[], total=0, hasMore=False)
+
+
+# CLI
+@click.command()
+@click.option("--port", default=3001, help="Port to listen on for HTTP")
+@click.option(
+    "--log-level",
+    default="INFO",
+    help="Logging level (DEBUG, INFO, WARNING, ERROR, CRITICAL)",
+)
+def main(port: int, log_level: str) -> int:
+    """Run the MCP Everything Server."""
+    logging.basicConfig(
+        level=getattr(logging, log_level.upper()),
+        format="%(asctime)s - %(name)s - %(levelname)s - %(message)s",
+    )
+
+    logger.info(f"Starting MCP Everything Server on port {port}")
+    logger.info(f"Endpoint will be: http://localhost:{port}/mcp")
+
+    mcp.settings.port = port
+    mcp.run(transport="streamable-http")
+
+    return 0
+
+
+if __name__ == "__main__":
+    main()