Search tools, AI agents, MCP servers
Submit a tool

Search iThub

Search tools, categories and collections. Type at least two characters.

Explore all tools
↑ ↓ to explore resultsEnter to open · Esc to close

Development & DevOps

What is InnerSource? Principles, roles, patterns and how to start

InnerSource applies open source practices inside a company. Its principles, roles, InnerSource patterns, how it differs from open source and how to start.

iThub Insights cover: Explainer, What Is InnerSource?, with the iThub logo tile
On this page
  1. Where does the term InnerSource come from?
  2. How is InnerSource different from open source?
  3. What are the core principles of InnerSource?
  4. Who does what in an InnerSource project?
  5. What are InnerSource patterns?
  6. What does an InnerSource-ready repository need?
  7. Which tools support InnerSource?
  8. How do you measure InnerSource?
  9. How do you start an InnerSource program?
  10. Which companies use InnerSource?
  11. Why do InnerSource initiatives fail?
  12. Sources

InnerSource is the use of open source development practices for software that stays inside one organization. Code is visible to engineers across the company, any team can propose a change to another team's repository, and the owning team reviews and merges those contributions in the open.

The InnerSource Commons, the community that maintains most of the shared material on the subject, defines it as the application of open source practices and principles to software development within the enterprise (InnerSource Commons). This guide is written for engineering leaders deciding whether to try it and for the developers who will maintain or contribute to InnerSource projects.

Where does the term InnerSource come from?

The earliest use people cite is Tim O'Reilly's "Ask Tim" column of December 2000, titled "Open Source and OpenGL". Answering a reader, he mentioned that O'Reilly had worked with companies on what he called "inner sourcing", helping them "use open source development techniques within the corporation" even when they were not ready to publish their code (O'Reilly, archived copy).

The practice got a written method in the following decade. Andy Oram's Getting Started with InnerSource (O'Reilly, 2015) described how PayPal changed its tools and engineering culture to adopt it (InnerSource Commons). Practitioners formed the InnerSource Commons as a community the same year, and it was incorporated as the InnerSource Commons Foundation, a 501(c)(3) public charity, in February 2020 (InnerSource Commons).

The community writes the name as one word, InnerSource. You will also see "inner source" and the lowercase "innersource", which is the spelling GitHub uses in its documentation.

How is InnerSource different from open source?

InnerSource borrows the mechanics of open source: readable repositories, pull requests, issue trackers, written discussion and maintainers who review outside work. The differences are the boundary around the code and who pays for contributors' time.

Aspect InnerSource Open source Traditional internal development
Who can read the code Every employee, unless the code is sensitive Anyone The owning team, sometimes others on request
Who can propose changes Any internal team Anyone with an account The owning team only; others file tickets
Who decides what is merged The owning (host) team, through its trusted committers The project's maintainers The owning team's lead or manager
Who pays for contributors' time The employer, agreed with the contributor's manager Volunteers or sponsoring companies Not applicable, only the owner writes code
License None needed inside one legal entity; a group-wide agreement when several legal entities share code An open source license, such as one approved by the OSI None
Where discussion happens Internal issue trackers and chat that any employee can read Public issue trackers, mailing lists and forums Meetings, private chat and ticket queues
What blocks a dependent team Waiting for review of its own pull request Waiting for review, or maintaining a fork Waiting in the owning team's backlog

When review by the host team becomes the slowest step for contributing teams, it is a classic bottleneck; our guide to the theory of constraints shows how to find and relieve it.

Is sharing code the same as InnerSource?

No. Making repositories readable is a sensible first step, and GitHub's guidance for enterprises recommends making repositories visible to all employees unless they contain sensitive information (GitHub Docs). Read access on its own mostly produces copy and paste reuse: a team forks a library, patches it locally and drifts away from the original. InnerSource starts when the owning team commits to accepting changes. It documents how to contribute, reviews outside pull requests within an agreed time and keeps the code workable for people who are not on the team.

Do you need licenses inside the company?

Usually not. Inside a single company the employer already owns the code, so open source licenses and SPDX headers add nothing. Groups made of several legal entities are the exception: the InnerSource License pattern describes a reusable legal agreement for sharing code between subsidiaries. What every program does need is clear ownership of each repository and an agreement on who maintains a contribution after it is merged.

What are the core principles of InnerSource?

The InnerSource Commons learning path names four principles:

  • Openness. Projects are easy to find and join: the code, its README and a documented way to contribute are reachable without asking for access.
  • Transparency. Decisions, roadmaps and discussions are written where people outside the team can read them. A guest who can see why an earlier proposal was rejected writes a better one.
  • Prioritized mentorship. The host team treats helping contributors as part of its job rather than an interruption. The learning path asks trusted committers to put mentoring during a pull request ahead of release dates unless there is a very good reason not to.
  • Voluntary code contribution. Engineers contribute because the change helps them or their team, not because someone assigned them to another team's backlog. Voluntary does not mean unpaid or invisible: the contributor's manager still has to accept the time spent.

Who does what in an InnerSource project?

InnerSource Commons describes two teams and three roles. The host team owns the software and reviews contributions. The guest team needs a change and writes it.

  • Product owner (host team). Decides which features the host team is willing to accept, keeps the backlog readable and says early when a proposal does not fit, before a guest spends a week writing code.
  • Contributor (guest team). Writes the change, follows the contribution guidelines and opens the pull request. Experienced contributors open an issue first to agree on the approach.
  • Trusted committer (host team). Reviews contributions, sets quality standards, keeps the contribution guidelines current and helps contributors get their changes merged. The learning path also expects trusted committers to advocate for contributors with management and to watch the long-term health of the codebase (InnerSource Commons).

The role takes time away from feature work, and the learning path notes that some companies rotate it so off-duty committers can return to technical work. The Trusted Committer pattern adds a second lever: a guest who keeps returning with good contributions can be made a trusted committer, which spreads the review load beyond the original team.

What are InnerSource patterns?

InnerSource patterns are documented solutions to problems that keep recurring in InnerSource programs. The InnerSource Commons publishes them as a free book, InnerSource Patterns, with 26 patterns at the time of writing. Each one describes a problem, its context, a solution and known instances in real organizations, and the community treats a pattern as mature once members have reviewed it and at least one organization is known to use it.

The patterns most teams meet first:

  • 30 Day Warranty. The contributing team fixes bugs in its contribution for 30 days after the code reaches production, which answers the host team's reluctance to own code it did not write. The book notes that GitHub uses it internally with a six-week period.
  • Service vs. Library. When sharing a running service would leave two teams arguing over who is on call, they share the code as a library and each team operates its own deployment.
  • Standard Base Documentation. Contributors answer their own questions from standard files: README.md, CONTRIBUTING.md and COMMUNICATION.md.
  • InnerSource Portal. An intranet site indexes InnerSource projects so contributors can discover them.
  • Repository Activity Score. A score derived from repository statistics such as stars, watchers and forks orders the projects in a portal, so active ones appear first.
  • Start as an Experiment. A time-limited experiment with stated criteria lets managers approve the program without a permanent commitment.

Other entries include the Gig Marketplace, a Maturity Model and Transparent Cross-Team Decision Making using RFCs.

What does an InnerSource-ready repository need?

Most of the work happens in the repository, not in a tool purchase. A practical checklist:

  • A README written for outsiders. What the project does, how to run it locally and whether it is actively developed, frozen or being replaced. Longer docs can live in a separate site; see iThub's documentation tools.
  • CONTRIBUTING.md. How to build and test, what a good pull request looks like, which changes the team will not accept, how quickly reviews happen and who the trusted committers are. Mercado Libre names its trusted committers in this file, according to the patterns book.
  • A CODEOWNERS file. On GitHub, a CODEOWNERS file in .github/, the repository root or docs/ maps paths to people or teams, and those owners are requested for review automatically when a pull request changes their code. With the branch protection option "Require review from Code Owners", no change merges without the owning team's approval (GitHub Docs). GitLab has a comparable Code Owners feature.
  • Issue and pull request templates. Asking for the use case, the requesting team and the deadline saves a round of questions. Label a few issues for first-time contributors.
  • Written decisions. Architecture decision records or RFCs in the repository show guests why the code looks the way it does.
  • One communication channel That the host team reads every day, named in COMMUNICATION.md.

Then change the default: new repositories start with internal visibility, and private visibility is kept for code that is sensitive for a stated reason.

Which tools support InnerSource?

Any Git forge with pull requests can host InnerSource projects. What matters is an organization-wide visibility level, code ownership rules and a way to find projects.

  • GitHub Enterprise offers internal repositories, which any member of any organization in the enterprise can view. GitHub recommends a base permission of at least Read in each organization, and guest collaborators do not see internal repositories by default (GitHub Docs).
  • GitLab Self-Managed and GitLab Dedicated have an Internal visibility level for projects and groups. Any signed-in user can view and clone an internal project, while external users cannot (GitLab Docs).
  • Self-hosted forges Work as well. Gitea can set an organization to limited visibility, which shows it to every signed-in user, and Bosch ran a later phase of its program on Atlassian's Stash, now Bitbucket.
  • Backstage , the open source developer portal, keeps a software catalog of components and their owners. Ownership metadata lives in YAML files committed with the code and changes through the normal Git workflow (Backstage docs), which makes the catalog a natural base for an InnerSource portal.

For more options, browse the iThub categories for version control and code hosting and development and DevOps.

How do you measure InnerSource?

Measure whether contributions cross team boundaries and whether guests get answers. All of these can be pulled from a forge's API:

  • Cross-team contributions Merged pull requests whose author is not on the owning team, per repository and quarter.
  • Time to first response How long a guest's issue or pull request waits for a human reply. Long waits are the earliest warning sign.
  • Time to merge for guest pull requests, Compared with the host team's own. A wide gap means outside work is being pushed down the queue.
  • Repeat contributors Guests who come back for more changes, and how many become trusted committers.
  • Reuse How many teams depend on a component, from dependency graph data, and how many duplicate implementations get retired.

Use these numbers to find projects that need help, not to rank engineers: counting pull requests per person rewards trivial changes. The Cross-Team Project Valuation pattern shows how to present the value of shared work that does not map directly to revenue. Record a baseline before the pilot starts, or nobody will be able to tell whether anything changed.

How do you start an InnerSource program?

  1. Find a sponsor and call it an experiment. A director or VP who can protect maintainer time matters more than any tool. Set a time limit and entry criteria for projects, as the Start as an Experiment pattern recommends.
  2. Pick one or two pilot projects. Good candidates are shared components with a queue of requests from other teams, such as an internal SDK or a deployment pipeline library. Avoid code that is about to be rewritten.
  3. Name trusted committers and budget their time. Agree with their managers how much of the week goes to reviews and mentoring.
  4. Write the base documentation. README, CONTRIBUTING.md, COMMUNICATION.md, CODEOWNERS and templates go in before anything is announced.
  5. Set and publish the contribution rules: a review response target, a warranty period for contributed code and a way to settle disagreements.
  6. Make the pilots findable. Internal visibility, repository topics and an entry in the software catalog or portal.
  7. Recruit the first guests on purpose. Invite the teams that filed the most requests, pair each with a trusted committer for its first pull request and thank contributors in public, as the Praise Participants pattern suggests.
  8. Review at the end of the experiment. Compare the metrics with the baseline, ask guests and hosts what slowed them down, then expand, adjust or stop.

Which companies use InnerSource?

These examples are documented by the companies themselves or by the InnerSource Commons:

  • Robert Bosch. Bosch Internal Open Source (BIOS) began in 2009 as a three-year experiment with corporate funding, full-time community leaders and a review committee of engineering vice presidents. Senior management ended that funding in 2015. A grassroots initiative called Social Coding continued until 2018 on a smaller budget and provided shared Git platforms to business units (InnerSource Commons case study).
  • PayPal. Andy Oram's 2015 book documents its adoption, and the patterns book lists PayPal as a known instance of both the 30 Day Warranty and the Trusted Committer patterns.
  • SAP. Runs an InnerSource portal where projects register themselves with GitHub topics and are ordered by the Repository Activity Score, a pattern SAP proposed to the community in July 2020. SAP released the portal as an open source reference implementation, which Mercado Libre and Mercedes-Benz use, according to the patterns book.
  • Airbus. Its data science community started by collaborating on shared Python libraries, which led to a group-wide InnerSource scheme. Airbus tested SAP's portal, then moved to the Bazaar plugin for Backstage, which lists internal projects looking for contributors (patterns book).

The InnerSource Commons case studies page collects more, including Microsoft, Europace and Deutsche Bahn.

Why do InnerSource initiatives fail?

The same problems show up across case studies and pattern write-ups:

  • No time budget for maintainers. When trusted committers keep their full feature commitments, outside pull requests wait and guests fork. Review time belongs in sprint planning.
  • Code thrown over the wall. A guest team lands a large change and moves on, leaving the host team to support it. A warranty period and small pull requests limit the damage.
  • Missing documentation. Without a CONTRIBUTING file, every contribution starts with a meeting.
  • Review bottlenecks. A single trusted committer per repository turns into a queue. Rotate the role, promote repeat contributors and use CODEOWNERS to route reviews to more than one person.
  • Contributions that do not count. If performance reviews only credit work on a team's own roadmap, engineers stop spending time on other teams' code. Managers have to recognize that work explicitly.
  • Funding that disappears. A program that depends on one central budget can stop when the budget does, as Bosch's first phase did in 2015.

Sources

Frequently asked questions

Is InnerSource the same as insourcing?

No. Insourcing means bringing work that was outsourced back to in-house staff. InnerSource describes how internal teams collaborate on code and says nothing about who employs the developers, so contractors can take part as contributors too.

Is InnerSource only for large companies?

Most published case studies come from large organizations, because the problems InnerSource addresses grow with the number of teams. A company with two or three teams that talk every day gets little from a formal program, although habits such as contribution guidelines and code owners start paying off once more teams depend on shared code.

How does InnerSource relate to platform engineering?

Platform teams own shared services that many product teams depend on, which makes them natural InnerSource hosts. Instead of filing a ticket and waiting, a product team can propose the change itself, while the platform team stays responsible for review and operations. A developer portal helps both sides by showing what exists and who owns it.

Does InnerSource work in a monorepo?

Yes. A monorepo already gives everyone read access, so discovery is easier and ownership becomes the hard part. Per-directory code owners, required reviews from the owning team and contribution notes for each component do the job that separate repositories and their README files do elsewhere.

Do InnerSource contributions have to be code?

No. Bug reports with a reproduction, documentation fixes, test cases, design feedback and answers to other users' questions all count, and they are often how a guest earns enough trust for a larger code change to be accepted. Tracking only merged code undercounts real participation.

Do engineers need their manager's approval to contribute to another team's project?

For anything beyond a small fix, in practice yes, because the hours come out of their own team's plans. Some organizations formalize this with agreements on how much time a contributor may spend, which the InnerSource Patterns book calls Contracted Contributor, or with an internal marketplace that lists contribution opportunities with estimated effort.

Is InnerSource a step toward open sourcing a project?

It can be. A project that already has contribution guidelines, readable discussions and contributors from outside its team has much of what a public release needs. Publishing still requires choosing a license, a legal review and a check of the Git history for secrets or customer data, none of which InnerSource covers.

Tools in this article

  1. GitHubGit hosting with pull requests, Actions CI/CD, Copilot, Codespaces and a self-hosted Enterprise ServerVersion Control & Code Hosting73 views
  2. GitLabDevSecOps platform with Git repositories, merge requests, CI/CD, security scanning and planningVersion Control & Code Hosting38 views
  3. GiteaSelf-hosted Git service with code review, issues, package registries and GitHub Actions compatible CI/CDVersion Control & Code Hosting4 views
  4. BitbucketGit hosting and Pipelines from AtlassianVersion Control & Code Hosting20 views
  5. BackstageOpen source framework for building developer portals around a software catalog, templates and TechDocsPaaS & App Deployment7 views
Share this article

Facts link to their official sources. Product details can change after publication: report an error or read the editorial methodology.

Emanuel DE ALMEIDA

Written by

Emanuel DE ALMEIDA

Entrepreneur · IT consultant · Software creator

Geneva-based entrepreneur, IT consultant and creator of SuniPDF and SuniCapture. Emanuel shares practical IT, cybersecurity and software insights on iThub.

Keep reading

More guides and analysis from the iThub editors.

All insights
What is Boolean logic? Operators, truth tables and how IT teams use itWhat is Boolean logic? AND, OR, NOT, truth tables, De Morgan's laws, short-circuiting, SQL NULL and bitmasks, with tested examples for developers and IT.How to make software architecture decisions: a practical guideA practical guide to architecture decision making: spot the decisions that matter, weigh trade-offs, write an ADR, pick a process and know when to revisit.12 open source projects for developers to know in 202612 open source projects for developers in late 2026, from Ghostty and Tauri to Langfuse and Zerobyte, with licenses and latest releases checked.