For companies, not individuals. A software company holds the most connected record of work in any office: commits, reviews, tickets and the chat that explains them, each with a timestamp and a link to the next. This guide covers why that link matters, what has to come out first, and which published rules apply.
Last checked: 7 October 2026. Buyer terms and cited prices are quoted as published.
Most companies have scattered files. A software team has a chain where each step points to the one before it.
customer reports a problem
triaged, scoped, assigned
options argued, one chosen
the change, step by step
critique and correction
tests pass, ticket closed
A commit on its own shows what changed. The issue shows why. The review shows what a senior engineer thought was wrong with the first attempt, and the chat thread shows which options were rejected. Joined together, they record how a team reasons from a problem to a working answer. That is the shape micro1 describes when it lists project histories, QA processes and "decision-making patterns" among what it wants.
The buyers' own source lists read like a software company's tool stack: GitHub, GitLab and Bitbucket repositories "with history", Jira and Asana for work tracking, Slack and Microsoft Teams for chat, Confluence and Notion for documentation, and Zendesk and ServiceNow for support. Few other industries keep every one of these in machine-readable form with cross-references already in place. Our guide to codebases and git history covers repository exports in detail, and support tickets covers the customer end of the chain.
There is a further reason the combination is valued. Practitioners say raw data is the cheapest tier, that evaluations built on the data are worth roughly ten times raw, and that full training environments reach six to eight figures but need heavy engineering. A repository with a test suite and real bug histories is closer to an evaluation than most company data, because the tests define what "correct" means. Whether a buyer builds that, and whether you share in it, is a negotiation. The tiers are explained in raw data vs evaluations vs environments.
Software companies have no special size line. The general rules apply, plus one buyer that names software companies directly.
| Source | Published figure | Published eligibility or context |
|---|---|---|
| Mode | "$100K-$5M" per company | 20+ full-time US office employees; several years of records the company owns; US-based teams the strongest fit |
| micro1 | "$100K-$2M+ for approved data packages" | 30+ employees (referral posting: 30 to 200); mature operations, documented processes, modern software tools; primarily English; US prioritized, then other Western markets |
| Grepped | "$20K-$5M" | Any vertical; also pays individual professionals for expertise |
| Miro Advisory | Private codebases "$10K-$1M+"; operating datasets "$100K-$1M+" | Indicative ranges; lists businesses and software companies. We have no referral link with it. |
| Troveo (closure market) | About $5,000 per code repository; roughly $10,000 to $100,000 per archive deal | Figures Troveo cites in its guide for the shut-down startup market (troveo.ai). The market itself was covered by Forbes on 16 April 2026, and by Fast Company and Gizmodo |
Last checked: 7 October 2026. Sources: data.mode.inc, micro1.ai/data-partnerships and micro1.ai/company-referral, grepped.ai, miroadvisory.com, troveo.ai, and Forbes, "AI's New Training Data: Your Old Work Slacks And Emails" (16 April 2026). Ranges are published or reported figures, not offers or averages.
Read the closure-market figures with care. They describe startups that have stopped operating and are selling an archive once. A running company is selling a scoped copy while it keeps building the same product, which changes both what it can include and what it should ask about resale. The company ranges at the top of the table are published across all sellers; a "+" or "up to" marks the top of the market, not a typical deal. To test your own size, systems and location against these rules, use the eligibility checker.
Version control remembers everything. A file deleted years ago is still in the history you export.
API keys, passwords and tokens anywhere in the commit history, including .env files committed once and removed. Scan everything and rotate what you find.
Test fixtures copied from production, logs, database dumps and screenshots in issues. These are often the hardest to spot.
Vendored libraries and open-source code under their own licenses, and code written by contractors without a clear IP assignment.
Custom integrations or features built under contracts that give the customer the IP, or that restrict how their materials are used.
Direct messages, HR and leadership channels. Public engineering channels carry most of the value with far less exposure.
Infrastructure repositories and incident notes that describe how to reach your systems. Leave out what you would not hand to a stranger.
Chat history raises its own questions, from employee notice to whether a channel was ever understood to be private; Slack and Teams messages covers them. Buyers describe their own privacy steps: micro1 says sensitive and confidential information is scrubbed and originals deleted after processing, and Mode says it de-identifies before onward delivery. Those steps come after you hand data over. Secrets and code you do not own should never leave in the first place.
Software has no single sector regulator, so the legal questions come from ownership, contracts and personal data. Raise these by name:
Questions for your lawyer: Do our customer agreements allow use of support content after de-identification? Does any contributor lack a signed assignment? Which third-party code must be removed? What ownership and non-infringement warranties can we safely give?
General information, not legal advice. Talk to your own lawyer before you sign.
The company, its systems and its decisions are invented to show the method.
Example Software Inc. is a made-up B2B software company with 48 full-time employees in a US office. It has 22 GitHub repositories with seven years of history, Jira for issues, Slack for chat, Confluence for design documents and Zendesk for support.
Published rules against its facts. Mode's 20+ full-time US office line is met. micro1's 30+ line is met, and 48 falls inside its referral posting's 30 to 200 band. Grepped lists any vertical. Miro Advisory lists software companies. Meeting a rule is not acceptance.
| Item | Decision | Reason |
|---|---|---|
| Core product repository with full history | In, after cleanup | Secrets scan across all commits; one vendored GPL directory removed on counsel's advice. |
| Jira projects linked to commits | In | Customer names in issue text replaced with consistent placeholders. |
| Public engineering Slack channels | In, after review | Staff notified first; threads quoting customer data removed. |
| Design documents and incident postmortems | In, partly | Postmortems describing live infrastructure access left out. |
| Zendesk tickets linked to bug fixes | In, after PII removal | Data processing agreements checked first; two enterprise customers excluded entirely. |
| Three custom-integration repositories | Out | Contracts assign that code to the customers. |
| Direct messages and leadership channels | Out | Private by expectation and low value to the chain. |
The company would send a manifest (repository count, years, issue volume, link coverage) and a few cleaned samples, approach more than one buyer, and compare offers on exclusivity and use as well as price.
Three invented companies show how size changes both eligibility and the shape of the history. Illustrative, not an offer.
The repository may be the main asset. The chain around it is often short: issues are informal, review is light, and decisions happen in direct messages or calls. A buyer will weigh the code on its own.
If the company is winding down rather than running, the closure-market figures Troveo cites are the closer comparison.
This is the size where issue tracking, code review and public channels tend to be consistent enough that ticket-to-release chains hold together across years.
The main work is cleaning history and reading customer contracts, which at this size are often still standard terms rather than negotiated ones.
More repositories and more history, but also enterprise customers with negotiated contracts and a security team that will want to review every export.
Security commitments made to customers, for example in questionnaires or audit reports, may limit what can leave your environment at all.
Most problems in a code sale are found too late, after an export has already left. These are the common ones.
vendor or third_party folder is someone else's code under someone else's license.One decision sits with the founders or leadership rather than engineering: whether licensing a copy of the codebase is compatible with how the company competes. The buyer itself may never compete with you, but the scope-of-use and resale terms decide where the copy can go next. Engineers can tell you what is in the history. Only leadership can decide what the company is willing to have outside it.
Staff deserve a decision too. Every commit carries an author name and email address, and every review and chat message is attributed to a person. That metadata is what lets a buyer follow a decision from one tool to the next, so stripping it entirely lowers the value. Consistent pseudonyms, where each engineer becomes the same placeholder in Git, Jira and Slack, keep the links while removing the names. Tell your team what is being licensed and how they will appear in it before the export, not after. Former employees are in the history as well, and they will not hear about it from a team meeting.
Generic questions every seller should check, weighted to a company that keeps selling the product its code describes.
Exclusive, time-limited or open? Can you license the same history to a second buyer? See exclusivity and resale rights.
Can the buyer pass your code on to downstream customers, and to whom?
Training only, evaluation, or building environments? Is any use excluded?
What do you warrant about ownership and third-party code, and is it limited to your knowledge?
Who pays if a secret or customer record slips through? Is liability capped and time-limited?
When is the copy deleted? Is payment one-off, by milestone, or on acceptance?
Practitioners cite 60 to 90 days to close, through NDA, review, agreement, export, de-identification and acceptance.
Independent site. Some links are referral links: if your company signs with a buyer through them, the buyer may pay us a fee. You are not charged, and we never see your data.
We are not a partner, agent or representative of any buyer, and we cannot promise acceptance, amounts or timing. The buyer runs discovery, contracts, export, de-identification and payment. See all terms side by side on buyer programs compared.A repository alone shows what changed. Linked to the ticket, the review and the chat thread, it shows why it changed and who decided. micro1 lists project histories, QA processes and decision-making patterns among what it wants, and buyers list GitHub, GitLab and Bitbucket repositories with history.
Miro Advisory publishes an indicative $10K-$1M+ for private codebases. In the market for shut-down startups, Troveo cites about $5,000 per code repository. Company-level ranges include Mode $100K-$5M and micro1 $100K-$2M+ for approved data packages. None of these is an offer.
Mode lists 20+ full-time US office employees. micro1 lists 30+ employees, with its referral posting saying 30 to 200, plus documented processes and modern software tools. Grepped lists any vertical. As published, checked 7 October 2026.
You can only license what you own or have the right to license. Vendored third-party and open-source code carries its own license terms, so identify it and ask your lawyer whether it must be removed before export.
Yes. Scan the full commit history, not only the current branch, for credentials, keys and customer data in fixtures or logs. Rotate anything found, and consider leaving out repositories whose history is hard to clean.
For Mode (20+ full-time US office employees) and micro1 (30+ employees), yes by their published lines. Grepped publishes no size line, and Miro Advisory lists software companies with an indicative $10K-$1M+ range for private codebases. As published, checked 7 October 2026.
Usually, if it is your own code. Tests define what correct behavior means, which is part of why practitioners describe evaluations as worth more than raw data. Ask the buyer whether it needs runnable code and how that changes the scope and price.