Zip Files in S3 with Python and the S3zipper API

If you write Python, the S3zipper API gives you a zipped archive of any S3 prefix in a handful of requests — no boto3 multipart plumbing, no temporary files. This is the complete flow.

1. Get a token

import requests

BASE = "https://api.s3zipper.com"

token = requests.post(
    BASE + "/gentoken",
    data={"userKey": "YOUR_USERKEY", "userSecret": "YOUR_USERSECRET"},
).json()["token"]

headers = {"Authorization": f"Bearer {token}"}

Store the key and secret as environment variables; the token expires after a day, so request a fresh one per run.

2. Start the zip job

start = requests.post(
    BASE + "/v1/zipstart",
    headers=headers,
    json={
        "awsKey": "AKIA...",
        "awsSecret": "...",
        "awsBucket": "my-bucket",
        "awsRegion": "us-east-1",
        "filePaths": ["my-bucket/reports/2026"],
        "zipFileName": "reports-2026.zip",
    },
).json()
print(start["size"])  # e.g. "181 MB"

filePaths are full S3 keys: a prefix means "everything under this folder". The response also contains a taskUUID that identifies the job.

3. Poll until it finishes

import time

while True:
    state = requests.post(BASE + "/v1/zipstate", headers=headers, json=start).json()
    if state["State"] in ("SUCCESS", "FAILURE"):
        break
    time.sleep(2)

if state["State"] == "FAILURE":
    raise RuntimeError(state.get("Error"))

4. Get the download URL

result = requests.post(BASE + "/v1/zipresult", headers=headers, json=start).json()
url = result["results"][0]
print(url)

# download it
with requests.get(url, stream=True) as r:
    r.raise_for_status()
    with open("reports-2026.zip", "wb") as f:
        for chunk in r.iter_content(1 << 20):
            f.write(chunk)

The URL is a presigned S3 link, valid for seven days by default. Requesting the result again returns the same link, so it is safe to share until it expires.

Notes for production use

Reuse one token for many jobs in the same process; it is valid for 24 hours.
Poll with a small back-off on large archives rather than a tight loop.
Handle FAILURE explicitly — a bad key, unreadable path or plan limit all come back there with a reason.
Add resultsEmail if you would rather receive the link than poll.