What Is an Internal Developer Platform?

An internal developer platform, or IDP, is self-service tooling that a company builds for its own engineers. A developer asks for a database, an environment or a deployment, and gets it without waiting on another team. The people who build it are usually called a platform engineering team.

The phrase to hold on to is "paved paths". The goal is a golden path: one supported way to do a common job, such as getting a new service to production, that is easier to follow than to avoid. The tooling is secondary. If nobody uses the path, you have built a very tidy shelf.

The CNCF platforms white paper defines a platform for cloud native computing as "an integrated collection of capabilities defined and presented according to the needs of the platform's users". It lists thirteen capability domains. You would not build all thirteen. Here are the ones most teams start with.

• A web portal to provision and see what exists, and APIs or a CLI so the same actions can be automated.
• Golden path templates and documentation, so a new service starts from a known good shape.
• Automation to build, test, deliver and verify services.
• Development environments that match production closely enough to trust.
• Identity, secret management and security services, so nobody invents their own.
• Data services such as managed databases, and messaging or event services.

Notice what is not on that list: a specific product. An internal developer platform is an outcome assembled from parts. Kubernetes, Terraform, a portal and a CI system are common ingredients, and infrastructure as code is usually one of them. If that idea is new to you, start with infrastructure as code for beginners.

The best recent numbers come from the CNCF and SlashData Technology Radar, published in March 2026. It surveyed more than 400 professional developers in the fourth quarter of 2025. These are three separate findings, not slices of one pie, so do not add them up.

Read the middle row carefully. Plenty of organisations get platform-like results from teams sharing the work, without a dedicated team. It is a survey of developers, so it leans towards organisations that already use CNCF projects, and a small business will not see itself in the average.

Backstage is the name you will hear most. Its own documentation calls it an open source framework for building developer portals, and it is a CNCF Incubation project. It gives you a software catalogue, software templates for new projects and TechDocs for documentation. It is a front door and a catalogue. It does not run your servers.

That sentence is the honest cost of the whole category. The docs give the team four jobs: run the deployment, drive adoption, encode best practice into templates, and evangelise the portal. That is a permanent job, not a project you finish.

This is a worked example with assumed inputs, so swap in your own. Say a platform team is two engineers. Assume each costs $120,000 a year fully loaded, and a working year is 220 days. The team costs $240,000 a year. Assume the platform saves each application team two engineer-days a month by removing waits and repeated setup.

Nineteen teams. With five engineers in total, you will never get there. The sum is crude. It ignores the time saved by the platform team itself, and fewer outages matter too. But it shows the shape. A dedicated platform pays for itself by serving many teams, and a small company has one.

The break-even sum assumes people use what you build. That is the usual point of failure. The CNCF white paper warns that platform teams which rely on top-down mandates to achieve adoption are almost certain to find resistance and resentment. It says teams should treat their platforms like products and develop them together with users.

In practice that means three unglamorous habits. Pick one or two partner teams and build the first golden path with them, not for them. Ask what slows them down before you choose a tool. Then measure something, such as the time from a new repository to a running service, and check it falls. The paper also advises including product managers from the start, which is a cost the sum above leaves out.

A fair counter-view: the sum can also understate the case. Value is not only days saved. A platform that stops one bad production change, or shortens one outage, can be worth more than a year of saved setup. The white paper says good product management depends on measuring performance, both in numbers and in what users tell you. So keep a running record. If after a year you cannot show that teams ship faster or wait less, the honest answer is to shrink the platform, or to stop building it.

• Build a real internal developer platform when many teams provision similar things in different ways and you can staff a team to own it as a product.
• Share the work, the model 41% of surveyed organisations report, when a few teams can agree on conventions and a rota, without a dedicated team.
• Use a hosting platform that already provides the self-service part when the real problem is waiting for someone, not a shortage of tooling.

People often ask how an internal developer platform differs from a PaaS, and how platform engineering differs from DevOps. A PaaS is a product you use as it comes. An internal developer platform is tooling you assemble for your own workflows. DevOps is a set of practices. Platform engineering is the team that builds the tooling that makes those practices easy to follow.

Helm7 is one of those hosting platforms. It gives a small team the self-service outcome: a dashboard, a CLI, an API and an MCP server for the app, Postgres, auth, storage and domains together. The Pro plan is $29 a month, or $348 a year, against the $240,000 team in the sum above. That is not a like-for-like comparison, and it is not meant to be. It is the price of not needing the platform team.

It also stops somewhere. If you have dozens of teams and want your own golden paths, an internal developer platform on Kubernetes with a portal is the right tool, and Helm7 is not that. Where the two meet is access control. Helm7 gives an agent or a person a scoped key with a spend cap, and the platform page shows how the pieces fit. Prices are on the pricing page.