Datamule Pricing Update

Datamule's SEC Archive has been up for a bit. We now serve more data per year than the SEC.

Costs are stable, so we're cutting prices by 10x for dataset and object storage downloads. R2 download costs have been reduced from $10 per million to $1 per million. Cost per S3 GB downloaded has been reduced from $10/GB to $1/GB.

What this actually means

Download every 10-K filing: $0.20
Download every filing (~20 TB of data uncompressed): $19
Download an up to date XBRL dataset in parquet form: $19
Download an up to date insider transactions dataset: $0.30
Download every XML file in the SEC flattened and stored in parquet form (190GB): $190

Note: this is usage costs. Subscription costs are fixed. Also, you can query the corpus using SQL for much less.

What is Datamule

Datamule is a project to make tasks done by LLMs ~100000x cheaper. SEC data is our training set. That's why our products are so cheap, and why we've opensourced most of our code (1000+ GitHub stars).

We are entirely bootstrapped. Our early infrastructure was enabled via startup credits from: AWS, CF, and GCP. We are specifically grateful to: George, Melissa Kargiannakis, and Christopher Rotas.

How to use Datamule

Go to dashboard.
Buy $1-$5 in credits.
Try out the Filings Archive.
If the experience is enjoyable, try out other products.

What products does Datamule have (10/16/2026)

Every SEC Filing since 1993. Either in original SGML, or parsed into constituent files and stored in a tar file.
XBRL. Query or download every XBRL file in the SEC Corpus.
Insider Transactions, Proxy Voting Records, Institutional transactions, and 100+ tables derived from SEC XML files. Query or download these.
Metadata on SEC Filings, both as reported by the SEC, and extracted from the filings themselves. Also includes Datamule's infrastructures metadata, for easy auditing.
SEC Filing Notifications. Instant. Websocket, webhooks, or by email.

Archives & datasets updated instantly. SQL queries update daily. This list of products will shortly expand dramatically.

Pricing

Datamule has usage costs, which are a nominal fee to use a product. For example, downloading a lot of files. Usage fees allows us to have no rate limits. Datamule also has subscription costs. These are generally ~$20/month to use a product. Subscription fees let us share more products with users. You can cancel subscriptions right away, and retain access to the beginning of the next cycle.

See up to date pricing.

Who should use Datamule

Professors who want to do research projects with students, but are prevented due to licensing. (We have no restrictions on commercial or noncommercial use)
Day traders who like the intersection of low cost and high quality.
Companies that want to avoid lock in dependency (our code is opensource).
Startups training new AI models.

Who should not use Datamule

People who expect a lot of customer service for $3.95.
People who want to host their own SEC infrastructure. Please host your own! Here is some code to get you started.

I've noticed a bug, and want to report it

Great! Please post an issue on Github, or email me: [email protected].

AI generated text is fine - a lot of non native English speakers as well as non technical people use datamule. If you are interfacing through Claude that's okay!

Here is how I like to get emails:

Hi John,

I have an issue with the object transfer to my S3. [X] happened. I've attached my Claude's logs below in error.md. I've asked Claude to make the error.md as concise as possible, with exact code that had errors.

Best,
Yours truly

This helps me know what parts of the product you or your agent are having trouble with. The conciseness lets me process it quickly, and get back to other work.

Note: I do not like to get fully AI generated emails. Or AI text mixed in without disclosure. If you are not considerate of my time, I will not respond.

Caveat Emptor

Buyer is responsible for checking if the data works for them.
Billing has a few quirks right now that may overcharge or undercharge user by ~10% per query. This is something that is common across many very large companies that I will not name here. They often solve it by obfuscating billing. We don't do that.
datamulehub is a newer package. It's a bit less polished than datamule-python. For example, when you download a lot of filings, using many workers, it might error for a few. Those filings exist, just try downloading them again. This will be fixed.

"Help! I accidentally spent my whole balance on a query."

This happens occasionally to PhD students using vibe coding to interact with datamule. Along with the price reduction update, we've also added the ability to restrict spend on keys, defaulting to $20. This replaces the older system of simply putting as much money into your account, as you feel comfortable with. You can still do that too.

Note: When I was a predoc at Berkeley, I accidentally spent $3,000 on a very large company's cloud. This was not a fun experience. I was not supposed to be charged anything, as I was on a student free credit special -- I still ended up paying ~$300. I have setup Datamule's billing such that this does not happen.

There is no free tier

We've found that charging a nominal fee makes our userbase much nicer. This keeps our limited bandwidth focused on model training and new datasets.