AI Training Data · Code Datasets
Source code repositories for training AI coding agents
Real production code, licensed for AI training.
Octobit8 licenses complete, production-grade software repositories to AI labs and companies building coding agents—with working tests, version history, documentation and clean rights to train on.
01 / The problem
Public code isn’t enough
- 01
Scraped open-source code carries licence and attribution risk you can’t fully trace
- 02
Public repositories are dominated by tutorials, toy projects and abandoned forks
- 03
Few public codebases show how real business software is structured end to end
- 04
Coding agents need working builds and tests to learn from—not just source files
02 / What’s included
Repositories built like real software
Every repository is a working application our engineers built and shipped—not generated or scraped code.
Complete production applications
Full, working codebases for real business software—not snippets, tutorials or isolated files.
Version history
Commit history and messages that show how features, fixes and refactors evolved over time.
Tests and build setup
Test suites, build scripts and CI configuration so agents can run code and verify their changes.
Documentation and architecture
READMEs, setup guides and architecture notes that give agents the context senior engineers rely on.
Multi-layer systems
Frontend, backend, database, integrations and infrastructure code in the same repository.
Clean, licensable code
Documented rights to license each repository for AI training, with secrets and personal data removed.
03 / Catalogue
Code across many domains
The catalogue spans a wide range of industries and application types—below are a few examples. Tell us the domains you need, and the manifest lists stack, size, test coverage and history for each repository.
Healthcare
- Hospital and clinic management
- Diagnostics and lab systems
- Pharmacy management
- Patient engagement bots
EdTech
- School and college management
- Learning management systems
- Student and parent portals
- Assessment tools
Travel & Hospitality
- Hotel property management
- Restaurant POS and operations
- Travel booking and agency tools
- Channel management
AI systems
- RAG pipelines
- Multi-agent systems
- WhatsApp and web chatbots
- LLM integrations and evaluation
Examples only—ask us about other industries, languages and application types
04 / Use cases
How AI teams use it
Supervised fine-tuning
Train models on realistic, well-structured code written to production standards.
Reinforcement learning environments
Use runnable repositories with tests as environments where agents attempt tasks and get verifiable rewards.
Agent task generation
Create SWE-bench-style tasks—bug fixes, features, refactors—from real code and real history.
Evaluation and benchmarks
Build private benchmarks on code your models have never seen in public training data.
05 / How it works
From enquiry to delivery
- 01
Tell us what you need
Domains, languages, repository size, and how you plan to use the data.
- 02
Review the catalogue
We share a manifest with stack, size, test coverage and history for each repository, plus samples under NDA.
- 03
Agree the licence
A commercial licence for AI training, scoped to your use—exclusive or non-exclusive.
- 04
Secure delivery
Repositories delivered to your cloud storage or private Git hosting.
06 / Licensing
Documented rights, a commercial licence for AI training, and no secrets or personal data in the code.
FAQ
Frequently asked questions
Who owns the code, and can we use it to train commercial models?
Does the code contain personal or customer data?
Can we see the code before buying?
Do you offer exclusive licences?
Training a coding agent? Let’s talk data.
Request the catalogue