OpenAI Agents SDK FastAPI Starter with Vercel Sandbox

Minimal FastAPI app that runs the OpenAI Agents SDK with Vercel Sandbox on Vercel's Python runtime. Each request spins up an isolated microVM, gives the agent shell access to analyze data, and tears it down when done.

Prerequisites

Python 3.12+ (Vercel default is 3.12).
A Vercel account and Vercel CLI (npm i -g vercel).
An OpenAI or Vercel AI Gateway API key.

Environment variables

Variable	Required	Description
`OPENAI_API_KEY`	Yes	OpenAI or Vercel AI Gateway API key.
`VERCEL_TOKEN`	Yes	Vercel access token. Create one at https://vercel.com/account/tokens.
`VERCEL_TEAM_ID`	Yes	Your Vercel team ID (starts with `team_`). Found under Team Settings > General.
`VERCEL_PROJECT_ID`	Yes	Your Vercel project ID (starts with `prj_`). Found under Project Settings > General.
`OPENAI_DEFAULT_MODEL`	No	Default model when the request body omits `model`. Falls back to `gpt-4.1-mini`.
`OPENAI_BASE_URL`	No	When using AI Gateway, set this to `https://ai-gateway.vercel.sh/v1`

Setup

Copy the example env file and fill in your values:

cp .env.example .env.local

Then edit .env.local with your keys:

OPENAI_API_KEY=sk-...
VERCEL_TOKEN=your_access_token
VERCEL_TEAM_ID=team_xxx
VERCEL_PROJECT_ID=prj_xxx

Local development

uv sync
uv run uvicorn app:app --reload --host 127.0.0.1 --port 8000

Open http://127.0.0.1:8000 to use the interactive demo. The agent has shell access to a sandbox with sample sales data (sales.csv).

API endpoints:

GET /api/health returns {"status": "ok", "openai_configured": true}
POST /api/run with {"input": "Which region grew the most?"} runs the sandbox agent

Deploy to Vercel

vercel

Vercel detects app.py and the app ASGI instance. Dependencies come from pyproject.toml.

After deploying, make sure OPENAI_API_KEY, VERCEL_TOKEN, VERCEL_TEAM_ID, and VERCEL_PROJECT_ID are set under Project Settings > Environment Variables.

Timeouts

Sandbox creation and agent runs can take several seconds. Heavy workloads may need Fluid Compute or Vercel Workflow for durable steps.

How it works

Each POST /api/run creates a fresh Vercel Sandbox microVM with sample data.
A SandboxAgent with Shell capability receives the user's prompt.
The agent writes and runs shell commands inside the sandbox to answer the question.
The sandbox is torn down after the response is returned.

AI Gateway

If you want to use the Vercel AI Gateway with the OpenAI Agents SDK, there are three simple changes that are needed.

Create a new AI Gateway API key, and set it as your OPENAI_API_KEY environment variable under Project Settings > Environment Variables.
Set the OPENAI_BASE_URL environment variable to https://ai-gateway.vercel.sh/v1 under Project Settings > Environment Variables.
Ensure that the provider prefix on your model ID is not being stripped by setting the model_provider=MultiProvider(openai_prefix_mode="model_id") in your RunConfig as shown in the example.

License

MIT (match your org's policy when publishing).