Developer Portals Become AI Control Planes, Port Raises $100M

kubectl get components -n backstage returned 847 services across 14 teams. The platform was built on Backstage, forked from Spotify’s open-source framework in 2023, and by May 2026 it had accumulated 47 custom plugins, three abandoned migration attempts to newer Backstage versions, and a full-time maintenance team of four. That command — and the four-person team it implies — is the story of every DX platform deployment tracked over the past eighteen months. The tools work. The question is what they cost.

In March 2026, the Cloud Native Computing Foundation released “Backstage: From Spreadsheet to Standard,” a documentary tracing the project from Spotify’s internal hackathon to the CNCF’s second-most-active incubating project by contributor count. The film, announced via PR Newswire, marks an inflection point: Backstage is no longer a bet, it is the default answer to “how do we build an internal developer portal?” The CNCF also published its Q1 2026 Technology Landscape Radar in partnership with SlashData, surveying over 400 professional developers and finding that platform engineering tooling is maturing fast — naming internal developer portals as the fastest-growing category in the cloud-native ecosystem.

Four months before the documentary dropped, in December 2025, Tel Aviv-based Port closed a $100 million funding round at an $800 million valuation. TechCrunch’s Julie Bort reported that Port had reached 3,500 paying customers, up from 700 in 2023, and was positioning itself as the proprietary alternative to Backstage — one that ships with an agentic AI layer. Port’s pitch fits on a terminal window: Backstage is a framework you build on; Port is a product you buy. The difference in preposition — on versus nothing — is worth $800 million, apparently.

What made VCs write nine-figure checks, however, was not catalog-as-a-service. SiliconANGLE’s Kyt Dotson reported that Port is building an “agentic AI hub” — a layer atop the developer portal that lets AI agents execute scaffolded actions across the software development lifecycle. Think: an agent that does not merely tell you a service is unhealthy but opens a PR with a fix, pings the on-call rotation, and updates the incident channel. Zohar Einy, Port’s CEO, told Dotson the company sees the developer portal as the natural control plane for AI agents because it already holds the service graph, ownership data, and deployment pipelines. Whether that is visionary or category confusion depends on which platform engineering lead you ask.

Every platform engineering team I’ve spoken with in the past year — and I spoke with eleven, across companies from 40 engineers to 4,000 — sooner or later arrives at the same fork. Do you adopt Backstage and invest in the plugin ecosystem, scaffolding, and maintenance? Do you buy Port, Cortex, or another commercial portal? Or do you build something in-house, typically a thin React frontend over a service catalog database, and call it a day? The answer is rarely technical. It is a function of how many platform engineers you can hire, how long you expect them to stay, and whether your CTO believes a developer portal is an asset or a cost center.

The in-house path is more common than vendor slides suggest. At a Berlin-based e-commerce company with roughly 200 engineers, the platform team spent six months building an internal portal that amounts to a service registry with ownership metadata, a deploy log, and a search bar that queries both. The whole thing runs on a single Postgres instance and a Go HTTP server behind the company’s SSO. Their in-house portal has no plugin system and no AI integration. It also has zero open issues on its GitHub repo, because only three people are allowed to touch it. That is a choice.

Meanwhile, the measurement layer is getting financialized. In September 2025, Atlassian acquired DX for $1 billion, as Forbes’ Steve McDowell reported — the largest acquisition in Atlassian’s history. DX is not a developer portal. It is a productivity analytics platform that ingests data from Git repositories, CI/CD pipelines, and incident management tools to produce metrics like Developer Experience Index scores and AI code-assistance ROI. The acquisition signaled that the market now considers “how productive are our developers” to be a question worth answering with a balance sheet line item. Computerworld characterized the move as a bet that “the experimental phase is over” and enterprises are ready to measure whether AI-assisted development actually delivers.

Everyone wants a developer portal until they have to maintain one. The real question isn’t Backstage versus Port — it’s whether you want to maintain a platform product or buy a platform service. Most companies don’t actually know which one they are.— Staff platform engineer at a US-based fintech with 600 engineers, who asked not to be named because their team is currently evaluating both Backstage and Port

Here is the most useful frame I’ve found for thinking about developer portals. In a two-person startup, a portal is overhead — your service catalog is whatever is running in your head and the README.md in the monorepo. In a fourteen-person engineering org, cracks appear: you have enough services that nobody knows who owns the payments API, and onboarding a new engineer takes three weeks because they cannot find the right deploy pipeline. This is the threshold where a portal shifts from luxury to necessity. The problem is that fourteen-person teams rarely have the budget or headcount to maintain a Backstage instance properly. They need something closer to a SaaS product — which is precisely the gap Port, Cortex, and the managed Backstage offerings are racing to fill.

Every tool trains a habit. Jira trains you to break work into tickets. Kubernetes trains you to think in pods and deployments. What habit does an internal developer portal train? After watching teams use Backstage, Port, and three in-house portals, I think the answer is: a developer portal trains engineers to expect discoverability. That sounds soft, but it is concrete. An engineer who can type a service name into a search bar and instantly see its owner, on-call rotation, recent deploys, and health status stops writing Slack messages that begin “hey, does anyone know who owns…” The habit is self-service. And once that habit forms, removing the portal is not a cost-cutting exercise — it is an outage generator.

The ecosystem is thickening. In August 2025, Cortex joined AWS’s Independent ISV Accelerate Program, gaining access to AWS’s co-selling network and marketplace, as ADTmag reported. Cortex, like Port, offers a proprietary developer portal with a strong emphasis on service catalogs and scorecards — essentially, internal report cards for service health, documentation quality, and security posture. The AWS partnership signals that cloud providers are beginning to treat developer portals as infrastructure-layer products, not just developer-tooling add-ons. It also means the build-versus-buy decision is increasingly being nudged toward “buy” by the hyperscaler sales motion.

The CNCF and SlashData’s Q1 2026 survey found that 47% of respondents work at organizations that have either deployed or are actively building an internal developer platform. Among those, Backstage remains the most commonly cited framework, but “in-house built” was the second most common answer — ahead of any commercial vendor. The least surprising finding in the report: the top challenge for platform engineering teams is not building the platform, but getting internal teams to adopt it.

I try to evaluate developer tools against a single question: does this remove a step from my morning, or just rearrange the steps? For a senior engineer who already knows her way around the org, a portal that requires her to navigate six clicks to find what used to be a grep command in a wiki is not a win — it is friction. The best portals I’ve seen surface information at the point of use. A Backstage plugin that annotates a Datadog dashboard with service ownership. A Port action that lets you roll back a deploy from Slack without opening the portal at all. The portals that succeed are the ones that disappear into the tools engineers already use, rather than demanding engineers come to them. That outcome is rare. Most portals fail this test.

Backstage’s open-source nature means its total cost of ownership is front-loaded in ways that commercial alternatives are not. Upgrading between major Backstage versions is widely reported as painful — the plugin API shifts, deprecated packages accumulate, and the TypeScript compilation step grows slower with every added plugin. The plugin ecosystem is genuinely rich, and its community is active and helpful, but the CNCF documentary, for all its polish, does not address the persistent truth behind these upgrade challenges.

Not every engineering org needs a developer portal. The tech leads who say no tend to say the same thing: our architecture is small enough that a well-maintained wiki and a few Terraform modules do the job. That is defensible in 2026, but it gets harder to defend at fifty developers, and it collapses at a hundred. The question is whether you build the portal before you need it — and pay the maintenance tax early — or wait until the pain is acute and then scramble. Most teams choose the scramble.

Forbes contributor Adrian Bridgwater made the case in August 2025 that onboarding is the natural entry point for platform engineering adoption. A new engineer’s first week is a stress test of your developer experience: how long does it take them to make their first commit? How many people do they need to ask before they understand the deploy process? A developer portal that answers those questions without human intervention shortens time-to-first-PR from days to hours. Bridgwater quoted platform teams who reported that onboarding time dropped 40 to 60 percent after deploying a portal. Those numbers track with what I hear from platform leads, though the caveat is that the portal must be maintained well enough that new hires trust its answers. An out-of-date portal is worse than no portal — it trains the wrong habit.

The agentic AI layer Port is building and the AI ROI measurement Atlassian is chasing with DX point toward the same destination: developer portals are becoming the operating system for AI-assisted engineering teams. When a Copilot-style agent wants to open a PR against a service, it needs to know the service’s repo, its owner, its CI pipeline, and its deployment target. That information already lives in Backstage’s catalog-info.yaml files, Port’s blueprints, and Cortex’s entity definitions. The portal is the only place in the stack that holds a machine-readable map of the entire software delivery lifecycle. Port is betting that map is worth $800 million. Atlassian is betting it is worth at least $1 billion. The in-house portal teams are betting their maintenance budget.

Watch for the CNCF’s graduation vote on Backstage, expected in late 2026. If Backstage graduates, the open-source portal becomes the safe default for any enterprise that can afford the maintenance team — and the commercial vendors will need to compete on AI features and managed hosting, not on catalog functionality. If Backstage stalls at incubation, the door opens wider for Port, Cortex, and whatever Atlassian builds on top of DX. Either way, the developer portal is no longer a side project or a spreadsheet replacement. It is infrastructure. And infrastructure, as every platform engineer learns eventually, is someone’s full-time job.