Python goes generally available on Vercel with native SDKs and Fluid compute

Vercel has promoted Python to general availability, making it a first-class language and runtime on a platform that has until now been centred on JavaScript and frontend frameworks. According to the Vercel blog, the effort began ten months ago with three goals in mind: idiomatic Python SDKs, a shared workflow for Python backends and JavaScript frontends, and compatibility with the wider Python ecosystem.

Runtime and framework support

The Vercel changelog states that FastAPI, Flask and Django applications are recognised and configured automatically, with framework-specific handling for serving static assets through the Vercel CDN; the accompanying blog post adds FastHTML to the list of detected frameworks.

Under the hood, Vercel runs CPython on Linux and reads dependency files managed by uv, including pyproject.toml. Request-driven applications run on Fluid compute, which Vercel says manages concurrent requests and scales with traffic. For long-running workers that pull jobs from queues or databases, Vercel Daemons keeps containers alive along with the application's Python packages and system dependencies.

Native Python SDKs

Vercel has built a set of Python SDKs for its own infrastructure, designed around language conventions such as type annotations, async/await and context managers. The changelog lists six:

• Sandbox: an isolated Linux VM created on demand, aimed at applications and agents that need to run untrusted code.
• AI SDK for Python (beta): generates or streams model responses and calls tools; installable with uv add ai.
• Workflow SDK for Python (beta): persists execution state and retries failed workflow steps.
• Queues (beta): send and receive messages for background work.
• Connect: obtains credentials for calling external APIs.
• Blob: uploads and retrieves files in object storage.

The SDKs can be used individually or combined inside a single Python application. To show how the pieces fit together, the Vercel team built a collaborative notebook application: a Vite frontend paired with a FastAPI backend and PostgreSQL. Each user's code runs in a Sandbox hosting Jupyter, published notebooks are rendered to HTML and stored in Blob, Queues push updates to connected browsers over WebSockets, and an AI assistant runs through the AI SDK inside Workflows. The source code is available on GitHub.

One project for frontend and backend

Vercel Services ties a Python backend and a JavaScript frontend into a single project, with shared preview deployments, atomic deploys and rollbacks, per-service logs and private service bindings. In the documented vercel.toml setup, a web service rooted at frontend/ and an API service exposing a FastAPI entrypoint such as main:app sit side by side, with rewrite rules sending /api/* requests to FastAPI routes unchanged and everything else to the frontend. Both services start together under vercel dev and deploy from the same repository.

Vercel's investment in Python

The blog post also details Vercel's work inside the Python community. The Gel Data team, led by the post's author and Elvis Pranskevichus, joined Vercel last December specifically to build Python support. Vercel sponsors the Python Software Foundation and is funding a year of CPython maintenance by contributor Serhiy Storchaka, who has authored more than 600 commits to CPython's main branch so far in 2026, according to the post. The company also sponsored PyCon US and EuroPython this year, and it published PEP 827, a proposal for type manipulation features that, if accepted, could let frameworks such as FastAPI expose simpler APIs with less boilerplate.

Why it matters

Most AI and machine learning code is written in Python, but Vercel's deployment story has historically been optimised for JavaScript frontends. With a generally available runtime and native SDKs for sandboxes, queues, durable workflows and AI model calls, Vercel is positioning itself as a place to ship AI agents and Python APIs next to the frontends that consume them — one repository, one preview pipeline, one rollback. For teams currently running FastAPI or Django services elsewhere, the pitch is a managed runtime with serverless-style scaling plus the developer experience Next.js users have long had. The open question is how this competes with platforms where Python was first-class from day one; Vercel says it will keep improving SDK maturity, runtime performance and workflows for both human developers and coding agents.