01What the programmes publish for software companies
| micro1 | Mode | Mercor | |
|---|---|---|---|
| Range | Tiers $100K–500K to $2M+ | $100K–$5M | $50K–$500K (software listing) |
| Minimum size | 30+ full-time staff | 30+ | 15–500 |
| Location | US first; strong UK demand | US strongest; others case by case | Anywhere |
| What it wants | Code, project histories, internal processes | Jira, Confluence, Slack and other systems | "The private code the engineering team has built, and the history behind it" |
Each programme's own pages, checked 30–31 October 2026. Ranges, not offers.
Where these records usually live
Code
- GitHub
- GitLab
- Bitbucket
Tickets and planning
- Jira
- Linear
Docs and chat
- Confluence
- Notion
- Slack
Examples of common tools, to help you find your records. Product names and logos belong to their owners; no affiliation or endorsement.
Mercor also lists SaaS and subscription companies (product databases, warehouse and analytics code; 50–1,000 staff with a data team), AI product companies (real usage data; from 5 staff) and cybersecurity companies, each with its own range. See Mercor's ranges by business type.
02Why the history matters
A codebase on its own is useful. The history around it is what's rare: commits with messages, code reviews, the tickets that drove each change, and the incidents that led to fixes. That record shows how a team actually reasons about software. Bring the Jira and ticketing data along with the repo.
03Three checks before you license code
- Open-source licences. Your repo almost certainly includes third-party open-source code. Some licences carry conditions on redistribution, so list them and ask the programme how it treats vendored code.
- Secrets and credentials. API keys, passwords and tokens hide in old commits. Scan the full history and rotate anything you find before any transfer.
- Client-owned code. If you build software for clients, their contracts may give them the code. Leave it out unless they agree.
Customer data in test fixtures, logs and database dumps needs the same care as any personal data. See UK GDPR.
04Before you apply
- Inventory repos, years of history and who owns each.
- Run a secrets scan across all history.
- List third-party and open-source components.
- Read the licence's scope, exclusivity and confidentiality terms with a lawyer. This isn't legal advice.
05Check which programmes you're likely to match
Our free fit check asks about your size, location, records and tools. The programme decides.
06Questions people ask
Can I sell my company's source code to AI companies?
You can license code your company owns. Mercor and micro1 both publish programmes that take it. Check open-source licences, scrub secrets, and leave out code your clients own.
Why do AI labs want private code?
Public code is already widely used. Private codebases with their history show how real teams build, review and fix software, which is what coding models need.
Does licensing code mean giving it away?
No. You license a copy for training under agreed terms. Read the scope, exclusivity and confidentiality terms closely.
Sources
- data.mercor.com/refer (Software Company listing), checked 30 October 2026
- micro1.ai/data-partnerships, checked 30 October 2026
- data.mode.inc, checked 31 October 2026
