api.sh for Python

Official Python client for api.sh: public web data from around 1,000 websites as clean JSON: search engines, social media, e-commerce, real estate, jobs, travel, news and sports (Google, Amazon, TikTok, Zillow, LinkedIn and more). Every endpoint is a typed method.

pip install apish-sdk

from apish import Apish

apish = Apish()  # reads APISH_API_KEY; or Apish("ash_live_…")  (https://app.api.sh/keys)

profile = apish.tiktok.profile(username="nasa")
print(profile["followers"], profile.meta.credits_used, profile.meta.request_id)

results = apish.amazon.search(q="headphones", options={"raw": False})

Python 3.9+, sync (Apish) and asyncio (AsyncApish).

Calls

• apish.<platform>.<endpoint>(**params, options=...): one method per endpoint, with keyword-only typed parameters and the price in its docstring.
• apish.call(route, params, options): any route, e.g. apish.call("/reddit/v1/subreddit", {"subreddit": "python"}).
• Options: raw (False leaves out raw source blocks), fields, fresh, output ("csv" / "ndjson" returns text), timeout, max_retries.
• Results are dicts with .meta (request_id, credits_used, credits_remaining, status, seconds).
• Retries 502/503/504 and network errors with backoff (failed calls are never charged).

Live catalog: counts, prices and new platforms

The typed methods cover every endpoint that existed when this version was published. The client also reads the live catalog (free, no credits, cached for an hour), so newer platforms and endpoints work without an upgrade:

cat = apish.catalog()                      # cat["platform_count"], cat["endpoint_count"]: today's numbers
prices = apish.prices()                    # route -> credits per call, live
apish.new_site.search(q="shoes")           # a platform added after this release
apish.hacker_news.some_new_endpoint(id="1")  # an endpoint added after this release

for page in apish.paginate("/reddit/v1/subreddit", {"subreddit": "python"}, max_pages=5):
    print(len(page["posts"]))

from apish import ApishError
try:
    apish.zillow.property(url=url)
except ApishError as e:
    print(e.status, e.message, e.insufficient_credits, e.spend_limit, e.request_id)

Async

import asyncio
from apish import AsyncApish

async def main():
    async with AsyncApish() as apish:
        print((await apish.reddit.subreddit(subreddit="python"))["posts"][0]["title"])

asyncio.run(main())

Account (free)

apish.account.get()            # balance, plan, spend, limits
apish.account.usage(days=30)
apish.account.logs(status="error")
apish.account.result(request_id)   # a past response again, free, 30 days

AI agents

from apish.tools import openai_tools, anthropic_tools, run_tool
tools = openai_tools()             # apish_search + apish_call (or openai_tools(["/tiktok/v1/profile"]))
text = run_tool(apish, name, arguments)   # when the model calls a tool

# LangChain:  pip install "apish-sdk[langchain]"
from apish.langchain import apish_tools
tools = apish_tools(apish)

For Claude Desktop, Cursor and other MCP clients, use the MCP server.

Docs: https://api.sh/docs/sdks · License: MIT