Internal developer portal decisions tend to get made in Q4, when platform budgets for the next year are set, and they tend to get made on demos. That is a mistake. Backstage, Port and Cortex are often lined up as three versions of the same product, but they start from different assumptions about who does the work, who the portal is for and how you pay for it. Pick the wrong model and you either end up with a framework no one has time to maintain or a SaaS bill that grows faster than your engineering headcount.

This comparison covers what each one is, what it costs in money and engineering time, and which team shape each fits. Pricing below was checked against vendor pages in October 2026 and will change; verify before you sign anything.

What Each One Actually Is

Backstage: a framework, not a product

Backstage is an open source framework for building developer portals, created at Spotify and accepted into the CNCF Incubator in March 2022. The core pieces are the software catalog, software templates (the scaffolder), TechDocs and a plugin system that most of the ecosystem builds on.

The important word is framework. You run it, you assemble the plugins and you own upgrades. Backstage ships frequent releases, and the project has been in the middle of a major frontend change: v1.42.0 declared the New Frontend System adoption-ready, and v1.49.0, released in March 2026, brought it to a 1.0 release candidate and made it the default for new apps. If you adopted Backstage several years ago, you are likely carrying a migration on top of normal plugin upgrades.

Spotify also sells Spotify Portal, a commercial, fully managed SaaS built on the open source framework, which is now generally available. It is a distinct option from self-hosted Backstage and is worth pricing separately if you want the Backstage ecosystem without running it yourself.

Port: a configurable SaaS portal

Port is a commercial SaaS portal built around a user-defined data model. You define the entity types that matter to you — services, environments, clusters, cloud accounts, whatever your organization tracks — and Port ingests data into them from your integrations. Self-service actions and automations sit on top of that model.

Port publishes its pricing, which is unusual in this category. As of October 2026 its pricing page lists a Free plan for up to 15 seats and 10,000 entities with no time limit, Basic starting at $30 per seat per month, Standard starting at $40 per seat per month and custom-priced Enterprise. Basic and Standard are billed annually, and SSO is only listed on Standard and Enterprise — a detail that pushes most companies with an identity provider straight to the Standard tier.

Cortex: a service-quality platform with a catalog

Cortex is also commercial SaaS, but its center of gravity is different. Its catalog exists to support Scorecards, which grade services against standards for production readiness, security and reliability, and Initiatives, which turn those standards into tracked, deadline-driven work across teams. It also offers an Eng Intelligence module for delivery metrics.

Cortex does not publish prices. Its pricing page describes per-seat pricing that scales with organization size and asks you to contact sales.

The Real Cost Comparison

Backstage: free license, expensive headcount

The license costs nothing. The engineers do not. Backstage requires a team that knows TypeScript and React, can keep plugins working across upgrades and can build the integrations that turn the catalog from YAML files into something that reflects reality.

The common failure mode is not a bad launch. It is a good launch followed by a catalog that drifts because nobody maintains the catalog-info.yaml files, and an upgrade that breaks plugins owned by third parties. If you cannot dedicate engineers to the portal on an ongoing basis — not just for the build — Backstage’s free license is the most expensive option on this list.

Port: predictable per-seat spend

With published per-seat prices, Port is the easiest to model. A worked example from the published list prices: 200 engineers on Standard at the $40 per-seat starting price is $8,000 per month, or $96,000 per year, before any negotiation and assuming every engineer needs a seat. Compare that with the fully loaded cost of the engineers you would otherwise assign to running Backstage. For many mid-sized organizations, the SaaS is cheaper; for large ones with an existing platform team, it often is not.

Cortex: priced like a leadership tool

Without public pricing, the only honest statement is that you will need a quote. Budget for it as you would any per-seat engineering tool, and push hard on which seats actually need access: scorecards are often consumed by managers and service owners rather than every engineer.

Which Team Shape Each Fits

Choose Backstage if you have a funded platform team that will own the portal as a product, you need deep customization or homegrown integrations, and you would rather spend engineering time than license fees. Backstage is also the safest bet against vendor lock-in, and BackstageCon, co-located at KubeCon North America on November 9, is a good place to see how other teams run it at scale.

Choose Port if you want a working portal quickly, your requirements are mostly catalog plus self-service actions, and you do not have spare engineers to run a framework. The free tier is genuinely useful for a proof of concept against your real systems before any sales conversation.

Choose Cortex if your main problem is not discovery or self-service but standards: getting hundreds of services to production-readiness criteria and showing leadership measurable progress. Cortex treats the scorecard as the product, and the catalog as its foundation.

Consider a hybrid only if you already have Backstage working well. Some organizations keep Backstage for the catalog and documentation and add a scorecard tool for maturity tracking. Starting with two tools is rarely worth the integration cost.

Questions to Settle Before Any Demo

  1. Who is the primary user — the developer creating a service, the on-call engineer looking up ownership, or the VP asking which services fail the security baseline? Each answer points to a different tool.
  2. Where does catalog data come from? If ownership and dependencies must be hand-maintained in YAML, the catalog will rot regardless of vendor. Prefer automatic ingestion from your source control, cloud accounts and Kubernetes clusters.
  3. What is the self-service scope? Creating a repository from a template is easy everywhere. Provisioning a database with guardrails is where the tools diverge.
  4. What does exit look like? Ask how you get your data model and catalog out if you leave.

A portal is a means to an end, and our guide to building an internal developer platform covers the platform it sits on. Before you compare features, write down the one workflow you want to make faster and the metric that shows it worked. Then run that workflow end to end in a trial of each candidate. The tool that makes that single path work with the least ongoing effort from your team is the right one, regardless of how the feature matrices compare.

Sources: Backstage release notes and CNCF incubation announcement (March 2022); Port pricing page (checked October 2026); Cortex pricing page and documentation (checked October 2026).