Run code execution on the Grok API

The code execution tool lets Grok write and run Python in a sandbox for exact math, stats, and data work. xAI SDK name: code_execution. Responses / OpenAI-compatible name: code_interpreter. Vercel AI SDK: xai.tools.codeExecution(). The sandbox includes common libraries (NumPy, Pandas, Matplotlib, SciPy), has time and memory limits, and has no external network or durable filesystem between requests.

Enable the tool

curl https://api.x.ai/v1/responses \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $XAI_API_KEY" \
  -d '{
  "model": "grok-4.6",
  "input": [
    {"role": "user", "content": "Calculate the compound interest for $10,000 at 5% annually for 10 years"}
  ],
  "tools": [{"type": "code_interpreter"}]
}'

import os
from xai_sdk import Client
from xai_sdk.chat import user
from xai_sdk.tools import code_execution

client = Client(api_key=os.getenv("XAI_API_KEY"))
chat = client.chat.create(
    model="grok-4.6",
    tools=[code_execution()],
    include=["verbose_streaming"],
)
chat.append(user("Calculate the compound interest for $10,000 at 5% annually for 10 years"))

for response, chunk in chat.stream():
    for tool_call in chunk.tool_calls:
        print(f"Calling {tool_call.function.name}: {tool_call.function.arguments}")
    if chunk.content:
        print(chunk.content, end="")

Analyze an attached file

Attach a CSV (or other data file) with input_file / file_url and keep code_interpreter in tools. The model loads the file, runs analysis code in the sandbox, and returns the numbers. See Chat with files on the Grok API.

from xai_sdk.chat import user, file
from xai_sdk.tools import code_execution

chat = client.chat.create(model="grok-4.6", tools=[code_execution()])
chat.append(user(
    "Total revenue by product, average units by region, and the top product-region pair.",
    file(url="https://docs.x.ai/assets/api-examples/documents/sales-data.csv"),
))
print(chat.sample().content)

When it helps

• Exact numerical results instead of approximate prose
• Multi-step calculations that need intermediate values
• Prompt-supplied datasets you want summarized or plotted in code
• Checking a math result before you trust it

Prefer grok-4.6 for code generation. Keep temperature low (about 0.0–0.3) for arithmetic.

Pitfalls

• Expecting packages outside the sandbox set, or outbound HTTP from the runner.
• Assuming files written in one request still exist on the next request — context is temporary.
• Vague prompts ("analyze this") — name the metrics, formats, and success criteria.