Every dataset is rights-reviewed before it’s listed
Engineering data you can explain to your counsel.
We source, review, and license software engineering datasets for teams training coding models, agents, and evaluations — each with a written record of who owns it and what it may be used for.
Popular:Code & test-case pairsCode changes & PR metadataCode review & debuggingEngineering workflow traces
- name
- issue-to-fix-histories
- version
- 1.2.0 (fixed)
- legal_owner
- reviewed
- inbound_licenses
- recorded
- permitted_use
- fine-tuning, evaluation
- secrets_scan
- passed
- dedup
- exact duplicates checked
- review_scope
- documented
Data for each stage of building a coding model
Pre-training, agent training, private evaluation, and custom sourcing — each with documented provenance.
Code model pre-training data
Rights-reviewed repositories and engineering history for continued pre-training of code models.
02Coding agent training data
Linked issue-to-fix histories and review trails that teach agents how real changes get made.
03Evaluation and benchmark data
Held-out evaluation sets from licensed private sources, kept out of public corpora.
04Custom engineering data sourcing
Tell us the languages, domains, and volume you need — we source from vetted suppliers.
Dataset categories
Data that captures how software actually gets built
We focus on engineering data that is hard to find in public corpora — linked, executable, and licensed by people with the authority to do so.
- Code & test-case pairsFunctions or modules paired with the unit and integration tests that exercise them.
- Code changes & PR metadataDiffs, commit sequences, pull-request descriptions, and review outcomes.
- Code review & debuggingReviewer comments, requested changes, and step-by-step debugging sessions.
- Engineering workflow tracesAuthorized records of how engineers plan, implement, and verify changes across tools.
- Issue-to-fix historiesLinked bug reports, the code changes that resolved them, and the tests that confirm the fix.
- Testing & evaluation setsHeld-out task suites with executable checks for benchmarking coding systems.
How it works
A managed marketplace, run by people who check the details
- 01Tell us what you needRequirements — data type, languages, volume, intended use — or browse listed datasets.
- 02 · decisionWe source and reviewVetted suppliers; rights review and technical QA recorded.
- 03Evaluate a sampleA time-limited sample once we approve your request.
- 04Sign, pay, get accessWritten licence signed online; pay by card or bank transfer; read-only access to a private copy of the repository.
Who it’s for
Two sides, one standard of evidence
For AI companies
Data you can explain to your counsel
- Datasets described honestly — contents, checks performed, and known limitations
- Structured rights records: owner, licences, restrictions, and permitted uses
- Custom sourcing when the catalog doesn’t have what you need
- Preparation to your schema: de-duplication, scrubbing, splits
For data owners
License engineering history you already own
- You decide which uses you will license, and on what exclusivity
- We handle buyer qualification, negotiation support, and contracts
- Preparation help: secret scanning, PII removal, formatting
- Withdraw from new licensing at any time, subject to existing agreements
Data preparation services
From raw engineering history to training-ready data
Preparation is available to suppliers getting data ready for licensing, and to buyers who need data in a specific shape.
Scrubbing
Secret and credential detection, PII removal or pseudonymisation.
Normalisation
Consistent schemas across repositories, trackers, and review systems.
Filtering
Licence-based filtering and de-duplication; benchmark contamination screening on request.
Packaging
Train/eval splits and documentation, delivered as a private read-only repository.
Trust, provenance, and quality
No listing goes live on a single checkbox
Rights review, technical QA, and publication are tracked as separate decisions. A dataset is listed only when both reviews have passed, and every published version is fixed — a material change creates a new version that goes back through review.
We never infer that public or open-source code can be relicensed for any purpose. We check the licences, contributors, and agreements that apply to each source and each intended use, and we record what our review did and did not cover.
- Legal owner
- Who holds the rights
- Authority
- Submitter’s authority to license
- Inbound licences
- Open-source and third-party terms
- Restrictions
- Employer, client, NDA obligations
- Permitted uses
- Pre-training, fine-tuning, evaluation…
- Review scope
- What was — and wasn’t — reviewed
FAQ
Common questions
What kind of data do you provide?
Is the data legally cleared?
Do I upload my repositories?
How am I paid?
Open-source code is public — can’t it be used for anything?
Tell us what your models need to learn.
Share your requirements and we’ll come back with sources, samples, and a written proposal — or tell you honestly if we can’t help.