Roofing
$450,000
Roofing company
Exclusive · perpetual · full operating record
Your commits, tickets and postmortems show how software really gets maintained.
Public code has been read many times over. What's harder to come by is the private record behind a working product: the ticket, the pull request, the reviewer's objection, the fix that finally shipped and the incident that came after it. Teams building coding, support and operations agents need that trail. We take an anonymized summary of yours to several of them at once, and you choose the license.
Small commission on close. Nothing up front.
Usually held in
- GitHub
- GitLab
- Bitbucket
- Jira
- Linear
- Confluence
- Notion
- PagerDuty
- Sentry
- Zendesk
- Intercom
- Figma
- Slack
What buyers pay for in software and SaaS.
Buyers value records that show a decision and what came of it more than they value finished paperwork. These are the parts of your record they look for.
Code with its history
Private repositories with every commit, branch and message, showing how a real product changed over the years: migrations, refactors and the changes that got rolled back. Older stacks and less common languages are harder for buyers to find than another new web app.
Pull requests and reviews
The proposed change, the reviewer's line comments, the revision that followed and whether it merged. It's an experienced engineer saying what's wrong and why, which is what code-review models learn from.
Tickets linked to the fix
A bug report or feature request, the change that closed it and the tests that went from failing to passing. When the repo still builds and its tests run from a clean checkout, each pair can become a graded task.
Incidents and postmortems
The alert, the on-call timeline, the channel where it was worked, the root-cause write-up and the commit that fixed it. A trail from detection to repair is what on-call and reliability agents train and test on.
Support threads with outcomes
Multi-turn tickets and chats with the macro used, the escalation to engineering and the resolution. Customer names and their business details come out before delivery; the problem and how it was solved stay.
Specs, design docs and design files
RFCs, architecture decisions and product specs that can be traced to the code that implemented them, and design files paired with the front end that shipped. A decision followed all the way into code is useful to long-horizon agents.
Systems that hold it and what to watch for.
This is where the record usually lives. Where we've written a data guide for a system, its card links to it, with steps to find out how much you have.
- GitHubCode repositoriesRead the data guide
- GitLabCode repositoriesRead the data guide
- BitbucketCode repositories
- JiraIssue tracking and projectsRead the data guide
- LinearIssue tracking and projects
- ConfluenceDocs and knowledge base
- NotionDocs and knowledge base
- PagerDutyMonitoring and incidents
- SentryMonitoring and incidentsError events often carry your customers' end-user data, such as IP addresses, emails and request payloads. That data stays out.Sentry keeps error events for 30 days on the Developer plan and 90 days on Team, Business and Enterprise, so error details older than that are likely already gone. Export what you want to keep.
- ZendeskCustomer supportRead the data guide
- IntercomCustomer supportRead the data guide
- FigmaProduct designFigma Community files and purchased UI kits belong to their creators. Only designs your team made can be licensed.
- SlackTeam chatOn Slack's free plan you can only see the last 90 days of messages and files, and anything older than a year may be permanently deleted, so older history may already be gone.
Who buys it and what they do with it.
We describe buyers by category and don't publish their names. Each one is qualified and under NDA before seeing your summary, and when they bid you see every offer side by side, with exclusivity, term and payment next to the price.
Coding-agent and code-model builders
Teams training models to write and maintain software. Private, long-lived codebases teach them how production software is actually kept running, and private code hasn't already been folded into public training sets.
Evaluation and benchmark builders
Groups that measure coding agents need tasks no model has seen. Private issue-and-fix pairs that will never be published keep their value as test material far longer than public ones.
Simulated-workplace builders
Builders of practice environments for agents want one company's code, tickets, chat, docs and incidents all describing the same events, so an agent can work inside a world that hangs together.
Support and operations AI
Companies building support agents and on-call assistants. They need real ticket mixes, the policies agents followed, and incident timelines that end in a fix.
Design-to-code tools
Builders of UI generation tools want design files matched to the front-end code that was actually built from them.
What a package looks like to a buyer.
Before anyone signs, a buyer sees a summary shaped like this one. It is built from rounded ranges: no name, no website, no people, no addresses.
The numbers in this example are made up. Yours come from your intake and the review call, and you read the summary before any buyer does.
See the full processExample package · Software
Illustrative- Sector
- B2B SaaS, workflow software
- Region
- North America
- History
- 8+ years, continuous
- Record
- ~25 repos, ~50K commits, ~12K reviewed PRs, ~18K linked tickets
- Stack
- TypeScript and Go, plus an older Java service
- In scope
- First-party code, PRs and reviews, tickets, postmortems, support threads
- Left out
- Customer data, secrets, open-source and vendored code, DMs
What stays out by default.
Some records aren't yours to license, and some shouldn't leave the building. Intake and the review call bring these up early. The main ones for this industry are below.
Your customers' data
What customers put into your product is theirs, and you usually hold it as their processor or service provider. Under the GDPR a processor acts only on the customer's documented instructions, and one that decides its own purposes becomes a controller for that processing (Article 28(3)(a) and 28(10)). California limits a service provider to the purposes in its contract (Cal. Civ. Code 1798.140(ag); 11 CCR 7050). Production databases, app logs, error-tracker payloads and fixtures copied from real data stay out.
Secrets and infrastructure details
Keys, tokens, passwords, .env files, Terraform state, internal hostnames and IP addresses come out. A credential found anywhere in git history should be rotated before anything moves, because rewriting history doesn't remove it from existing clones.
Open-source and vendored code
Checked-in dependencies, copied snippets and GPL or AGPL files belong to their authors under their own licenses. Only code your company wrote and owns is in scope, and a license scan shows which files that is.
Code without clear title
Software written by an independent contractor generally isn't a work made for hire, so the company owns it only under a signed written assignment (17 U.S.C. 101 and 204(a)). Code an agency built for a client usually belongs to that client. Either stays out until the paperwork or the client's written consent is in place.
Security findings
Pentest reports, bug-bounty submissions, vulnerability details and anything about an unpatched issue stay out. A security postmortem is considered only once the issue is fixed and the write-up is cleaned.
Private and HR channels
Direct messages and HR, recruiting, compensation and social channels are excluded. In what remains, people's names are replaced before delivery under the license.
This is not legal advice. Whether a record can be licensed depends on your contracts, your software terms and how it was collected, and your counsel should review any license before you sign. The rules named on this page are linked under Sources.
Licenses in our network closed by buyers, not by us.
Recent licenses of operating records, closed by buyers in our network. They come from trades and services businesses rather than this industry, so read them as a sense of the market and nothing more.
Construction
$300,000
General contractor
Exclusive · 24-month term · full operating record
Solar
$275,000
Solar company
Exclusive · 18-month term · full operating record
Marketing
$110,000
Agency with global clients
Exclusive · 24-month term · full operating record
Licenses closed by buyers in our network, shown to describe the market. They are not deals thedata.broker brokered and not a forecast of what your record will bring. Every record is priced on its own.
Software and SaaS questions worth settling early.
More in Answers, or ask us on a call after your intake.
Usually, if employees wrote it as part of their jobs and contractors signed assignments. Under U.S. copyright law, code from an independent contractor generally belongs to the company only with a signed written assignment (17 U.S.C. 101 and 204(a)). Code an AI tool generated with little human input may not be protected by copyright at all, because copyright needs a human author (Thaler v. Perlmutter, D.C. Cir. 2025; U.S. Copyright Office, Copyright and Artificial Intelligence, Part 2). Client-owned code, vendored dependencies and open-source files are separate questions. Intake asks about each, and the license covers only what you can grant. This is not legal advice.
Only if you put it in scope. You can leave out core algorithms, pricing logic and security code, ask for evaluation-only terms, rule out direct competitors, and keep the license non-exclusive and time-limited. Buyers see an anonymized summary first, and nothing moves until you sign.
No, but that data stays out. Production data, logs and error events are excluded, and customer names and business details in tickets are removed before delivery. Check your customer contracts too: some promise no third-party disclosure or no use for AI training, and those customers' material is left out entirely. This is not legal advice.
Usually the public engineering, product and incident channels, with people's names replaced. DMs and HR or social channels stay out. How the data is exported matters too: Slack's API terms bar third-party apps from using API data to train large language models, so have counsel check your own Slack agreement before any export. This is not legal advice.
Yes. Federal rules treat developers of tax preparation software as tax return preparers (26 CFR 301.7216-1), and income tax return information is never licensed through us. Your code, tickets and engineering records can still be considered; the return data your product holds cannot. This is not legal advice.
We are paid a small commission out of a deal that closes. If nothing closes, you owe nothing. You keep ownership of your records and keep using them.
Sources
Primary sources for the rules this page names. Summaries above are not legal advice.
- GDPR, Regulation (EU) 2016/679: Article 28, processors (opens in a new tab)
- Cal. Civ. Code 1798.140: CCPA definitions, including service provider (opens in a new tab)
- 11 CCR 7050: CCPA regulations on service providers (opens in a new tab)
- 17 U.S.C. 101: work made for hire (opens in a new tab)
- 17 U.S.C. 204: transfers of copyright must be in writing (opens in a new tab)
- Thaler v. Perlmutter, No. 23-5233 (D.C. Cir. Mar. 18, 2025): copyright requires a human author (opens in a new tab)
- U.S. Copyright Office, Copyright and Artificial Intelligence, Part 2: Copyrightability (Jan. 2025): AI-generated material and human authorship (opens in a new tab)
- Slack API Terms of Service: use of API data (opens in a new tab)
- 26 CFR 301.7216-1: tax return preparers, including software developers (opens in a new tab)
Not quite your business?
See every industry we take on, the business functions buyers ask for, and which records need special handling.