Every figure on this page is pre.dev's own claim with its check date. The page follows the same method as every buyer profile on an independent reference on AI training data licensing: the primary page, the date and the absences are stated, and nothing is estimated.
Its scale claims are 10,000+ private codebases, 50B+ tokens and 2.5M+ commits, as of 2026-09-27, and its site terms name PREDEV CORP as the contracting entity (checked 2026-11-17).
Sources and check dates follow how DataPacts sources and dates every figure.
What pre.dev Licenses: Private Codebases With Full Git History
pre.dev takes one kind of data: private production codebases with their full git history, licensed directly from the teams that built them.
Its Labs page names two products for AI labs, and the record holds one input:
- private production codebases with full git history
- RL coding tasks built from real commits, the product pre.dev supplies
- company records other than code: none on record
pre.dev states 10,000+ private codebases, 50B+ tokens and 2.5M+ commits, "rounded down, as of 27 Sep 2026" (checked 2026-11-09). The counts show the size of its library, and they come from the Labs page, not from a third party. They do not show how many owners were paid or any amount paid.
Why git history is licensed and priced as its own unit, and how other acquirers compare, sit on the page about licensing source code and codebases to AI companies. That page owns the general questions, so this profile covers only what pre.dev itself states.
pre.dev Company Identity
The company behind pre.dev is Pre Dev Corp., a US company that pre.dev describes as supplying codebases and RL coding tasks to AI labs. The table lists what the Labs page and the site terms state, as of 2026-11-17.
| Attribute | Value | Source |
|---|---|---|
| Legal name on the Labs page | Pre Dev Corp. | pre.dev/labs |
| Brand | pre.dev | pre.dev/labs |
| Website | pre.dev | pre.dev |
| Contracting entity per site terms | "PREDEV CORP", with a Camden, Delaware address and Delaware governing law; terms last updated 2026-11-15 | pre.dev/terms_and_conditions |
| Public registry number | none on the primary pages checked 2026-11-09 | pre.dev/labs |
| Country | United States | pre.dev/labs |
| Type | RL-environment builder that licenses codebases | pre.dev/labs |
| Founded | not stated | n/a |
| Founders, headcount, funding | not recorded | n/a |
The Labs page and the site terms write the name differently, and no registry entry is on record here to settle the form. The founding date, the founders, the headcount and any funding are not stated on the pages checked, so none is given.
A code owner checks the name in the contract itself against a state registry before signing. The site terms cover use of the website, and whether they govern a code license is not stated on the pages checked.
pre.dev Program Terms and Payouts, Dated
pre.dev publishes no seller payout; its Labs page says codebases are priced per token, with one rate for source code and one for git history, quoted on a call (checked 2026-11-09). The table lists every term the pages state and each term they leave out, as of 2026-11-09.
| Term | What pre.dev states | Checked | Source |
|---|---|---|---|
| Program | pre.dev Labs codebases | 2026-11-09 | pre.dev/labs |
| Data wanted | private production codebases with full git history | 2026-11-09 | pre.dev/labs |
| Seller payout | not published | 2026-11-09 | pre.dev/labs |
| Pricing basis | per token, one rate for source and one for git history, quoted on a call; no public figure | 2026-11-09 | pre.dev/labs |
| Referral fee | none recorded | 2026-11-09 | pre.dev/labs |
| License | licensed directly from the teams that built the code; non-exclusive by default | 2026-11-09 | pre.dev/labs |
| Scope, term, deletion, warranties, payment trigger and timing | not published | 2026-11-09 | pre.dev/labs |
| Sign-up | as an individual or an organization | 2026-11-09 | pre.dev/code-licensing |
| Eligibility | exposure check, see below | 2026-11-09 | pre.dev/labs |
| Scale claims | 10,000+ codebases, 50B+ tokens, 2.5M+ commits, rounded down, as of 2026-09-27 | 2026-11-09 | pre.dev/labs |
The published terms settle three things: the license is non-exclusive by default, the code is licensed from the teams that built it, and a seller signs up as an individual or an organization. They leave the seller payout, the license scope, the term, deletion, warranties, the payment trigger and the payment timing to negotiation.
For code, pre.dev quotes the per-token rates on a call and publishes no figure, so a code owner has no published number to compare. Other programs publish rates or ranges, and the page of program terms compared, dated lines them up by date, with a check date on every row.
pre.dev's Exposure Check: What a Codebase Must Pass
pre.dev states that every codebase passes an exposure check: the repository must not be public, a fork, archived or a published package, and all branches and tags are included (checked 2026-11-09).
The checklist, as pre.dev states it:
- The repository is not public, not a fork, not archived and not a published package.
- All branches and tags are included.
- Secrets and personal data are removed and authors are pseudonymized.
- Vendored dependencies, binaries and build output are stripped.
Commit history carries author names, emails and sometimes secrets, which is why pre.dev states removal and pseudonymization. The owner's own pass before export is a separate step, covered under PII removal and de-identification before licensing.
Stripping vendored dependencies removes code the owner did not write. The license question behind that code, which can travel with a repository into a license, is covered under open-source license contamination when licensing a codebase.
How a pre.dev Code License Runs: Sign-Up, Samples Call, Delivery
pre.dev's pages describe three steps around a codebase license, from sign-up to delivery, and publish nothing on timing, acceptance or payment (checked 2026-11-09).
- Sign-up at pre.dev/code-licensing, as an individual or an organization.
- The Labs page describes a short form and a 30-minute call with the founders to review samples.
- Delivery of codebases through expiring signed links.
The Labs page addresses AI labs as its audience, so the form and the call are described as pre.dev states them, not as a step a code owner is promised.
The pages state no timeline, no written offer step, no acceptance or payment step and no deletion rule if no deal follows. Handling samples before any license is covered under proving data quality without sharing the data, and delivery formats and acceptance sit on their own page, linked below.
What pre.dev Builds From Licensed Code: RL Coding Tasks for AI Labs
pre.dev turns licensed codebases into RL coding tasks built from their real commits, and supplies both the tasks and the codebases to AI labs.
The labs are not named: no lab customer appears in the sources checked, so a code owner cannot tell which labs would hold the licensed code or the tasks built from it. What a task contains, a repository, an instruction and a check, is explained under RL environments for coding agents.
A non-exclusive license by default leaves the owner free to license the same code elsewhere. Who holds rights in an RL environment built from licensed code is a separate question that page answers.
pre.dev Reputation: The Record as of 2026-11-09
No complaint or incident involving pre.dev appears from a primary or named-person source as of 2026-11-09, and no court case or regulator action is on record.
A search for the brand with the word review returns pages about code review tools and forum posts, none about this entity, and forum threads are never evidence here.
The absence of a complaint is a statement about the sources checked, not a finding that none exists. The dated log of data buyer incidents holds no pre.dev entry.
Benefits, Risks, Downsides and Alternatives for a Code Owner Considering pre.dev
For a code owner, pre.dev offers a non-exclusive default and a stated exposure check, but no published payout, rate or contract terms, so the comparison runs against programs that publish numbers, and against not licensing at all.
Benefits
- Non-exclusive by default, so the code stays licensable elsewhere, which matters for price and lock-in under non-exclusive data licenses.
- Individuals such as solo developers sign up as well as organizations.
- A stated exposure check removes secrets and personal data and pseudonymizes authors before delivery.
Risks
- No published license terms beyond non-exclusivity, so scope, term, deletion, warranties and payment trigger are negotiated from zero.
- No public registry number, and two spellings of the name, so the entity is checked from the contract itself.
- The AI labs that receive the code are not named.
Downsides
- No published seller payout or rate, so there is no number to compare before the call.
- No referral terms recorded.
- Code only, with no route for other company records.
Alternatives (alphabetical, each one party's own claim checked 2026-11-09)
- Collinear AI: wants private repositories with pull requests and tickets; the seller keeps ownership; no amounts stated.
- Mercor Data Monetization: GitHub is among the supported tools; a wire typically follows 2 to 4 weeks after signing; no amounts stated.
- Pangea.ai: a marketplace that places code with AI labs; indicative USD 100 per million source tokens and USD 40 per million git-history tokens, as Pangea states (checked 2026-11-09).
- Patronus AI: production codebases with history, non-exclusive; states it does not resell code.
- Turing Project Lazarus: revenue share with nothing upfront; the license to Turing is exclusive for the first 6 months, during which Turing may license the record to more than one lab (checked 2026-11-17).
- Not licensing the code at all.
None of the options is better than another in general. Keeping the code is a complete outcome, as the page on alternatives to selling data sets out.
pre.dev Among Code Acquirers and Company Data Programs
pre.dev is one of several acquirers that take private codebases with history. It is also one of the company data partnership programs on record, and the links below lead to every comparable profile and to the pages that cover the steps around a code license.
- Every program and buyer: who buys AI training data
- Comparing options: choosing a buyer for your data
- Code profiles: Pangea.ai, Collinear AI, Patronus AI, Turing Project Lazarus, Mercor
- Builders: RL environment companies
- Before any license: who owns the code you want to sell and export rights for GitHub repositories
- After signing: delivering licensed data to an AI buyer, the license clause matrix and AI training data prices
More About pre.dev
pre.dev and Paid Developer Work
No program paying developers to write or review code for pre.dev appears in the sources checked as of 2026-11-09, and pre.dev's coding agent product is not covered here. Paid coding work for AI training is covered in AI training jobs and platforms compared.
Frequently Asked Questions About pre.dev
Does pre.dev buy code?
pre.dev licenses private production codebases with full git history from the teams that built them, non-exclusively by default; ownership is not stated to change (checked 2026-11-09).
How much does pre.dev pay for a codebase?
pre.dev publishes no seller payout; its Labs page says codebases are priced per token, one rate for source and one for git history, quoted on a call (checked 2026-11-09).
Does pre.dev take exclusive rights to the code?
Not by default: pre.dev states its licenses are non-exclusive by default; other terms are not published (checked 2026-11-09).
Which repositories does pre.dev reject?
Public, forked or archived repositories and published packages; vendored dependencies, binaries and build output are stripped (pre.dev, checked 2026-11-09).
Which company is behind pre.dev?
Pre Dev Corp., a US company; its site terms name PREDEV CORP, and no public registry number appears on the primary pages checked 2026-11-09.