AQUAVIEW MCP with ChatGPT (OpenAI Responses API)¶

This notebook shows how to give ChatGPT direct access to the AQUAVIEW MCP catalog using the OpenAI Responses API's native hosted-MCP tool — no proxy, no local server.

We'll ask ChatGPT the same question as in the Claude and Gemini notebooks:

Find Argo float profiles within 200 km of Hawaii from March 2026 where the float reached at least 1500 dbar pressure. Summarize the temperature range observed and give me a few download links.

ChatGPT will autonomously call AQUAVIEW's search_datasets, get_item, and possibly aggregate tools to answer.

Sources used: GADR (Global Argo Data Repository, NOAA NCEI)

Install¶

You only need the openai SDK.

In [ ]:

%pip install -U openai

%pip install -U openai

Setup¶

Set your OpenAI API key in the OPENAI_API_KEY environment variable.

In [ ]:

import os
from openai import OpenAI

assert os.environ.get("OPENAI_API_KEY"), "Set OPENAI_API_KEY in your environment"

client = OpenAI()

AQUAVIEW_TOOL = {
    "type": "mcp",
    "server_label": "aquaview",
    "server_url": "https://mcp.aquaview.org/mcp",
    "require_approval": "never",
}

MODEL = "gpt-5"

import os from openai import OpenAI assert os.environ.get("OPENAI_API_KEY"), "Set OPENAI_API_KEY in your environment" client = OpenAI() AQUAVIEW_TOOL = { "type": "mcp", "server_label": "aquaview", "server_url": "https://mcp.aquaview.org/mcp", "require_approval": "never", } MODEL = "gpt-5"

Ask the question¶

We pass the AQUAVIEW MCP tool in tools=[...] and the model can call it directly. ChatGPT will issue tool calls, the API will execute them against mcp.aquaview.org/mcp, and the model will integrate the results into its final answer.

In [ ]:

PROMPT = (
    "Find Argo float profiles within 200 km of Hawaii from March 2026 "
    "where the float reached at least 1500 dbar pressure. Summarize the "
    "temperature range observed and give me a few download links."
)

response = client.responses.create(
    model=MODEL,
    input=PROMPT,
    tools=[AQUAVIEW_TOOL],
)

print(f"status: {response.status}")
print(f"input  tokens: {response.usage.input_tokens}")
print(f"output tokens: {response.usage.output_tokens}")

PROMPT = ( "Find Argo float profiles within 200 km of Hawaii from March 2026 " "where the float reached at least 1500 dbar pressure. Summarize the " "temperature range observed and give me a few download links." ) response = client.responses.create( model=MODEL, input=PROMPT, tools=[AQUAVIEW_TOOL], ) print(f"status: {response.status}") print(f"input tokens: {response.usage.input_tokens}") print(f"output tokens: {response.usage.output_tokens}")

Inspect what ChatGPT did¶

The Responses API returns a sequence of output items including:

• mcp_list_tools — when the model first inspected the AQUAVIEW tool schema
• mcp_call — each tool call with input arguments and returned content
• message — the model's final response text

In [ ]:

from textwrap import shorten

for i, item in enumerate(response.output):
    item_type = getattr(item, "type", type(item).__name__)
    if item_type == "mcp_list_tools":
        print(f"[{i}] mcp_list_tools (server: {item.server_label})")
    elif item_type == "mcp_call":
        args = getattr(item, "arguments", {}) or {}
        out = getattr(item, "output", "") or ""
        preview = shorten(str(out).replace("\n", " "), width=240, placeholder="…")
        print(f"[{i}] mcp_call → aquaview.{item.name}({args})")
        print(f"    {preview}")
    elif item_type == "message":
        text = "\n".join(
            getattr(c, "text", "")
            for c in (item.content or [])
            if getattr(c, "type", "") == "output_text"
        )
        print(f"[{i}] message\n{text}\n")
    else:
        print(f"[{i}] {item_type}")

from textwrap import shorten for i, item in enumerate(response.output): item_type = getattr(item, "type", type(item).__name__) if item_type == "mcp_list_tools": print(f"[{i}] mcp_list_tools (server: {item.server_label})") elif item_type == "mcp_call": args = getattr(item, "arguments", {}) or {} out = getattr(item, "output", "") or "" preview = shorten(str(out).replace("\n", " "), width=240, placeholder="…") print(f"[{i}] mcp_call → aquaview.{item.name}({args})") print(f" {preview}") elif item_type == "message": text = "\n".join( getattr(c, "text", "") for c in (item.content or []) if getattr(c, "type", "") == "output_text" ) print(f"[{i}] message\n{text}\n") else: print(f"[{i}] {item_type}")

Typical trace for this prompt:

1. mcp_list_tools — model fetches AQUAVIEW's tool schemas
2. mcp_call → aquaview.search_datasets(collections='GADR', bbox='-161,18,-154,23', datetime='2026-03-01.../2026-03-31...', filter={...Pressure.max >= 1500...})
3. mcp_call → aquaview.get_item(collection='GADR', item_id='argo_jma_7900869') to fetch download URLs
4. message — ChatGPT's narrative summary citing the floats and their NCEI / Ifremer GDAC URLs

Render ChatGPT's final answer¶

In [ ]:

from IPython.display import Markdown, display

display(Markdown(response.output_text))

from IPython.display import Markdown, display display(Markdown(response.output_text))

Multi-turn — follow up¶

Pass the prior response.id as previous_response_id to keep the conversation context (and any prior MCP tool results) on the server side.

In [ ]:

followup = client.responses.create(
    model=MODEL,
    previous_response_id=response.id,
    input=(
        "For the deepest-diving float you found, give me its full lifetime "
        "trajectory bounding box and the direct NetCDF download URL."
    ),
    tools=[AQUAVIEW_TOOL],
)

display(Markdown(followup.output_text))

followup = client.responses.create( model=MODEL, previous_response_id=response.id, input=( "For the deepest-diving float you found, give me its full lifetime " "trajectory bounding box and the direct NetCDF download URL." ), tools=[AQUAVIEW_TOOL], ) display(Markdown(followup.output_text))

Streaming¶

For interactive UIs, stream the response so you can render the model's reasoning + tool calls + final text as they arrive.

In [ ]:

with client.responses.stream(
    model=MODEL,
    input=(
        "How many Argo profile observations are in the AQUAVIEW catalog "
        "from 2025? Use an aggregation, don't enumerate."
    ),
    tools=[AQUAVIEW_TOOL],
) as stream:
    for event in stream:
        et = getattr(event, "type", "")
        if et == "response.output_text.delta":
            print(event.delta, end="", flush=True)
        elif et == "response.mcp_call.in_progress":
            print(f"\n[tool] aquaview.{event.item.name}", flush=True)
    print()

with client.responses.stream( model=MODEL, input=( "How many Argo profile observations are in the AQUAVIEW catalog " "from 2025? Use an aggregation, don't enumerate." ), tools=[AQUAVIEW_TOOL], ) as stream: for event in stream: et = getattr(event, "type", "") if et == "response.output_text.delta": print(event.delta, end="", flush=True) elif et == "response.mcp_call.in_progress": print(f"\n[tool] aquaview.{event.item.name}", flush=True) print()

Where to go next¶

• Same notebook in Claude / Gemini: see 01-claude-mcp-agent.ipynb and 03-gemini-mcp-agent.ipynb — same prompt, three different agent runtimes.
• Other examples: see ../examples/ for 20+ more research questions, with prompts and transcripts.
• Tool reference: full parameter docs in ../docs/tools-reference.md.
• OpenAI Agents SDK (alternative): if you want multi-step orchestration with handoffs, the OpenAI Agents SDK supports MCP via MCPServerStreamableHttp — see the INSTALL.md snippet.