Async HTTP with httpx/aiohttp

httpx (recommended for FastAPI-era stacks) and aiohttp provide async HTTP clients with connection pooling, timeouts, and concurrent requests. Share one client instance per app lifecycle - not per request.

Recipe

import asyncio
import httpx
 
async def main() -> None:
    async with httpx.AsyncClient(timeout=10.0) as client:
        r = await client.get("https://httpbin.org/get")
        print(r.status_code)
 
asyncio.run(main())

When to reach for this:

• Fan-out API aggregation in FastAPI routes
• Webhook delivery and retries
• Service-to-service calls under load
• Streaming downloads/uploads
• Replacing blocking requests in async apps

Working Example

import asyncio
import httpx
 
URLS = [
    "https://httpbin.org/delay/0.1",
    "https://httpbin.org/status/200",
    "https://httpbin.org/uuid",
]
 
async def fetch_all(client: httpx.AsyncClient, urls: list[str]) -> list[int]:
    async def one(url: str) -> int:
        resp = await client.get(url)
        resp.raise_for_status()
        return resp.status_code
 
    return await asyncio.gather(*(one(u) for u in urls))
 
async def main() -> None:
    limits = httpx.Limits(max_connections=10, max_keepalive_connections=5)
    async with httpx.AsyncClient(timeout=5.0, limits=limits) as client:
        codes = await fetch_all(client, URLS)
        print(codes)
 
asyncio.run(main())

What this demonstrates:

• Shared AsyncClient reuses TCP connections (keep-alive)
• Limits caps concurrent connections per host
• raise_for_status turns 4xx/5xx into exceptions
• gather overlaps network waits on one loop

Deep Dive

httpx vs aiohttp

Client Lifecycle

• Create at app startup (FastAPI lifespan)
• Close on shutdown to flush pools
• One client per event loop thread

Gotchas

• New client per request - destroys pooling, exhausts sockets. Fix: app-scoped singleton.
• No timeouts - hung coroutines forever. Fix: default timeout on client and per-call override.
• Sync httpx in async route - blocks loop. Fix: AsyncClient only in coroutines.
• SSL/env proxy misconfig - works in dev, fails in prod. Fix: explicit trust_env and certs.
• Unbounded gather to slow APIs - overload peer and self. Fix: semaphore limit concurrency.

Alternatives

FAQs

Related

• Tasks & Gather - concurrent fetches
• Sync/Async Bridging - avoid sync requests
• Retries & Backoff - retry policy
• Asyncio Basics - loop fundamentals

Stack versions: This page was written for Python 3.14.0 (stable 3.14, maintenance 3.13), FastAPI 0.115+, Django 5.2, Flask 3.1, Pydantic 2, PyTorch 2.6+, pandas 2.2+, Polars 1.x, ruff 0.9+, and uv 0.6+.