Massive Flat Files

massive-flat-files

The bulk historical foundation. Any skill that needs more than a handful of ticker-days reads this first.

Why this exists

REST is great for "what's AAPL trading right now." It falls over when you need every trade for every name for every day across the last decade. Flat files give you that data as gzipped CSVs in an S3 bucket. No rate limits. No per-call overhead. Pull a year of minute aggregates for the entire US universe in a few minutes.

Access

The S3 bucket lives at s3://flatfiles/. Endpoint is https://files.polygon.io (legacy hostname, still works).

Credentials are NOT the REST API key. Flat files use S3-compatible auth with a separate access key ID and secret access key that you generate in your Massive dashboard under the Flat Files section. These two values are distinct from each other AND distinct from the REST API key you use for api.polygon.io. A common misconfiguration is to hardcode the REST API key as both aws_access_key_id and aws_secret_access_key, which returns 403 Forbidden on every S3 operation. If you hit 403s out of the gate, regenerate the dedicated S3 keys from the dashboard before suspecting an entitlement issue.

Entitlement gotcha (verified 2026-06-23): flat-files access is not automatically included with every paid plan despite what the marketing page implies. Test runs from a Stocks Business + Options Business + Benzinga add-on key returned 403 Forbidden on every list, head, and get operation against the bucket. The same key happily serves /v3/reference/tickers, /v2/aggs/grouped/..., snapshots, and options chains over REST. Verify flat-files entitlement on your key before building a workflow on it: hit s3.head_object(Bucket='flatfiles', Key='us_stocks_sip/day_aggs_v1/2026/06/2026-06-19.csv.gz') in a quick boto3 probe; a 403 means the key is not provisioned for the bucket and you need to contact support or use the REST fallback.

REST fallback for day aggregates: when flat-files are unavailable, GET /v2/aggs/grouped/locale/us/market/stocks/{date}?adjusted=true returns all US-listed stocks for that date in one call (~10,000 rows). Throughput is equivalent to one S3 day-bucket per trading day, just running over REST instead of S3. The schema is T / c / v / o / h / l / vw / n (capitalized fields) versus the flat-file schema of ticker / close / volume / etc (lowercase); your loader should normalize.

# These are the S3 keys from the Flat Files section of the dashboard,
# NOT the REST API key. They're two distinct values.
aws configure set aws_access_key_id ${MASSIVE_S3_ACCESS_KEY} --profile massive
aws configure set aws_secret_access_key ${MASSIVE_S3_SECRET_KEY} --profile massive
aws configure set endpoint_url https://files.polygon.io --profile massive

aws s3 ls s3://flatfiles/us_stocks_sip/day_aggs_v1/2026/06/ --profile massive

Python with boto3 or s3fs works the same way. The web-based File Browser at massive.com/dashboard/file-browser is the easiest way to discover paths.

Path layout

Files partition by asset class, data type, and date:

s3://flatfiles/{asset_class}/{data_type}/{yyyy}/{mm}/{yyyy-mm-dd}.csv.gz

Asset classes available:

• us_stocks_sip: US stocks
• us_options_opra: US options
• global_crypto: Crypto
• global_forex: FX
• us_indices: US indices

Data types per asset class:

• trades_v1: every tick
• quotes_v1: every NBBO update
• minute_aggs_v1: minute OHLCV
• day_aggs_v1: daily OHLCV

A single day of stocks trades is a few GB compressed. Plan disk and parallelism accordingly.

Reading a day file

import os
import pandas as pd

df = pd.read_csv(
    "s3://flatfiles/us_stocks_sip/day_aggs_v1/2026/06/2026-06-20.csv.gz",
    storage_options={
        # Dedicated S3 credentials from the Flat Files dashboard panel,
        # not the REST API key.
        "key": os.environ["MASSIVE_S3_ACCESS_KEY"],
        "secret": os.environ["MASSIVE_S3_SECRET_KEY"],
        "client_kwargs": {"endpoint_url": "https://files.polygon.io"},
    },
)

Schema is documented per data type at massive.com/docs/flat-files.

Parallelism

The bucket has no per-request rate limit. Workflows should fan out across days or symbols freely. Practical cap is your egress bandwidth and disk write speed, not anything on the Massive side.

from concurrent.futures import ThreadPoolExecutor

def fetch_day(date):
    return pd.read_csv(s3_path(date), storage_options=opts)

with ThreadPoolExecutor(max_workers=16) as pool:
    frames = list(pool.map(fetch_day, dates))

16-32 workers is a healthy default. Beyond that you saturate the pipe.

Cost

Included in every paid Massive plan at no extra charge. No egress fees from Massive. AWS egress to your local machine is free (Massive eats it).

If you're downloading TB-scale data to your own AWS account in a different region, you may see standard AWS inter-region transfer fees, but those are between you and AWS, not Massive.

When to use REST instead

• Looking up one or two names: REST is faster than spinning up an S3 client and decompressing a multi-GB file
• Real-time intraday state: flat files are end-of-day only
• Reference data (ticker list, exchange list, splits): REST endpoints exist for these and they're tiny

What lives here

• references/paths.md: full path layout reference for every asset class and data type
• references/schemas.md: column schemas per data type
• references/parallelism.md: download patterns for large pulls

What does NOT live here

REST patterns (massive-api-patterns) and live streams (massive-websockets).