Federal Docket Intelligence
The method, published before the numbers
This is where CaseMagic will publish original measurements of federal docket activity. There are none on it today, and the reason is worth more to you than a chart would be: we have not yet confirmed that the terms we read court data under allow us to publish aggregates of it. Until that is answered, this page carries the method and no figures.
Last reviewed . This date changes when a statement on the page changes, not when the page is served.
The licensing question, stated plainly
CaseMagic reads federal docket data through CourtListener and its RECAP archive, both run by the Free Law Project, under the API terms that come with an API token. Everything on this site that shows you a docket comes from there.
Two things are true at once, and confusing them is how this goes wrong:
- The underlying records are public and the facts in them are free. Facts are not copyrightable — Feist Publications v. Rural Telephone Service (1991) — and the work of a federal court is not subject to copyright under the government edicts doctrine, reaffirmed in Georgia v. Public.Resource.Org (2020). Nobody owns the fact that a motion was filed on a Tuesday.
- That is not the constraint. The constraint is the agreement we accepted in order to use a particular API at a particular rate. A public fact obtained under a contract that limits redistribution is still limited by that contract, whatever its copyright status.
So the open question is narrow and answerable: do the Free Law Project's API terms permit publishing aggregate statistics derived from CourtListener access? We have not had that answered, and the right way to answer it is to ask them rather than to read the terms hopefully. Bulk data for exactly this kind of analysis is offered separately from the API, which is itself a hint about which route is the intended one.
Nothing is published here until that is settled in writing. If the answer is no, this page will say so and stay empty of numbers, which is a perfectly good outcome for a page whose job is to be trustworthy.
What will be measured
All of it aggregate, none of it about an identifiable party, and each one chosen because a litigant or a small firm would change what they do if they knew it:
When dockets move
The distribution of docket activity across the days of the week and the hours of the day. Anyone who has waited on a filing has a folk theory about this. Nobody has published the measurement.
How long the quiet gaps run
The typical interval between entries in an active civil case, by stage. The number that answers “should I be worried that nothing has happened in five weeks”.
How districts differ
Practice varies by district in what reaches a docket and how quickly. A litigant in one district reading advice written for another is a real and invisible problem.
How fast a change becomes visible
The lag between a filing and its appearance in a public source. This one is partly about us, and publishing it is how we stay honest about what monitoring can and cannot promise.
The rules any number here will follow
- It comes from a real query. No estimate, no illustration, no figure carried over from somebody else's article.
- The query is shown. The filter, the window and the record count, next to the number, so it can be argued with.
- The measurement date is shown and is the date it was measured, not the date the page was served.
- No case is identifiable and no customer's matter is in the sample, ever.
- A number that turns out to be wrong is corrected in place and the correction is noted rather than quietly overwritten.
A free check reads the court source and then serves the same answer for 15 minutes. A watched case is re-checked about every 6 hours. Nothing here is real time, and every result carries the time it was last checked.
What is already published, and is not a statistic
Reference material about the courts themselves — official names, citation forms, the courthouses in each district and where each court's own website is — is fact about a public institution rather than an aggregate of case data, and it is on the federal district court pages. Docket data comes from CourtListener and its RECAP archive.
The things CaseMagic knows about itself — how often a watched case is re-read, how long a free check is cached — are on the product facts page, where a model that wants them does not have to infer them from marketing copy.
Meanwhile, the thing this page is about works
United States federal district courts — all 94 districts, including the territorial district courts. State, county and municipal courts are not supported yet.
No account needed to checkFederal courtsResults in seconds
Questions people ask before they start
Why are there no numbers on this page?
Because we have not yet confirmed that we are allowed to publish them. CaseMagic reads docket data through CourtListener under the Free Law Project’s API terms, and whether those terms permit publishing aggregate statistics derived from that access is a contract question we have not had answered. Publishing first and asking afterwards risks the data source the whole product runs on.
Is not court data public? Why is there a licensing question at all?
The two are different questions. Federal court records are public and the facts in them are not copyrightable — facts never are, under Feist v. Rural Telephone Service. What binds us is not copyright but the agreement we accepted to use a particular API. A public fact obtained under a contract that restricts redistribution is still restricted by that contract.
What will you publish once that is settled?
Measures that are useful and that nobody currently publishes for a general reader: how docket activity is distributed across the week, how long the gap between filings typically runs in an active case, how district practice differs in what appears on a docket and how fast. Aggregates only, and never anything identifying a party in a case.
Will you publish anything about your own customers’ cases?
No. Not in aggregate, not anonymised, not as an example. A customer’s matter is theirs, it never appears on a public page, and it is excluded by construction rather than by a promise — private paths are refused to every crawler by robots.txt and by a noindex header, and nothing in the account area is in the sitemap.
How will I know a number here is real?
Every figure we eventually publish will carry the query that produced it, the window it covers, the number of records it is drawn from, and the date it was measured. A statistic without those four things is decoration, and we would rather publish nothing than publish decoration.