docs: add pyhton agent tracing example (envoyproxy#1295)

**Description**

Adds an example Python agent that is fully instrumented to easily
showcase how tracing works.
This PR borrows the agent code from:
elastic/observability-examples#90

**Related Issues/PRs (if applicable)**

N/A

**Special notes for reviewers (if applicable)**

N/A

---------

Signed-off-by: Ignasi Barrera <[email protected]>
Signed-off-by: Hrushikesh Patil <[email protected]>

Author: nacx

examples/mcp/README.md

@@ -221,3 +221,26 @@ From here, you can start using the tools inside Claude.
   | LAX → SFO | 13/09 16:41 → 18:13 (1h 32m) | Economy | $129  | https://on.kiwi.com/F8u8Sc |
   | LAX → SFO | 13/09 16:41 → 18:13 (1h 32m) | Economy | $134  | https://on.kiwi.com/NBMEYO |
 ```
+
+## Tracing example with a Python agent
+
+The [agent.py](agent.py) file contains a basic Python agent example that evaluates the given prompt on the configured
+OpenAI compatible provider.
+
+Refer to the [cmd/aigw](../../cmd/aigw) directory README and Docker files for details and examples on
+how to start the different example OpenTelemetry compatible services.
+
+Once you have everything running you can start the agent by passing a prompt file or directly typing it
+into the terminal.
+
+```shell
+$ uv run --exact -q --env-file <environment file>> agent.py /path/to/prompt.txt
+```
+
+or
+
+```shell
+$ uv run --exact -q --env-file <environment file>> agent.py <<EOF
+> your prompt here
+EOF
+```

examples/mcp/agent.py

@@ -0,0 +1,85 @@
+# Copyright Envoy AI Gateway Authors
+# SPDX-License-Identifier: Apache-2.0
+# The full text of the Apache license is available in the LICENSE file at
+# the root of the repo.
+
+# run like this: uv run --exact -q --env-file .env agent.py
+#
+# Customizing the ".env" like:
+#
+# OPENAI_BASE_URL=http://localhost:1975/v1
+# OPENAI_API_KEY=unused
+# CHAT_MODEL=qwen3:4b
+#
+# MCP_URL=http://localhost:1975/mcp
+#
+# OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4318
+# OTEL_EXPORTER_OTLP_PROTOCOL=http/protobuf
+#
+# /// script
+# dependencies = [
+#     "openai-agents",
+#     "httpx",
+#     "mcp",
+#     "openinference-instrumentation-openai-agents",
+#     "opentelemetry-instrumentation-httpx",
+#     "openinference-instrumentation-mcp",
+# ]
+# ///
+
+from opentelemetry.instrumentation import auto_instrumentation
+
+# This must precede any other imports you want to instrument!
+auto_instrumentation.initialize()
+
+import argparse
+import asyncio
+import os
+import sys
+
+from agents import (
+    Agent,
+    OpenAIProvider,
+    RunConfig,
+    Runner,
+    Tool,
+)
+from agents.mcp import MCPServerStreamableHttp, MCPUtil
+
+
+async def run_agent(prompt: str, model_name: str, tools: list[Tool]):
+    model = OpenAIProvider(use_responses=False).get_model(model_name)
+    agent = Agent(name="Assistant", model=model, tools=tools)
+    result = await Runner.run(
+        starting_agent=agent,
+        input=prompt,
+        run_config=RunConfig(workflow_name="envoy-ai-gateway"),
+    )
+    print(result.final_output)
+
+
+async def main(prompt: str, model_name: str, mcp_url: str):
+    if not mcp_url:
+        await run_agent(prompt, model_name, [])
+        return
+
+    async with MCPServerStreamableHttp({"url": mcp_url,"timeout": 300.0},cache_tools_list=True) as server:
+        tools = await server.list_tools()
+        util = MCPUtil()
+        tools = [util.to_function_tool(tool, server, False) for tool in tools]
+        await run_agent(prompt, model_name, tools)
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser("Example Agent with Tools")
+    parser.add_argument("prompt", help="Prompt to be evaluated.", default=sys.stdin, type=argparse.FileType('r'), nargs='?')
+    parser.add_argument("--model", help="Model to use.", default=os.getenv("CHAT_MODEL"), type=str)
+    parser.add_argument("--mcp-url", help="MCP Server to connect to.", default=os.getenv("MCP_URL"), type=str)
+    args = parser.parse_args()
+    prompt = args.prompt.read()
+
+    print(f"Prompt: {prompt}")
+    print(f"Using model: {args.model}")
+    print(f"Using MCP URL: {args.mcp_url}")
+
+    asyncio.run(main(prompt, args.model, args.mcp_url))