Over the past twelve months, we have evaluated internal developer portals for multiple client engagements ranging from 40-person startups to 500-engineer enterprises. The question we kept hearing was the same: “Should we build with Backstage or buy something like Port or Cortex?”
After deploying all three in proof-of-concept environments and maintaining two of them in production, we have a clear answer — and it is not the one most blog posts will give you. The reality is more nuanced than “Backstage is free” or “just buy a SaaS.” This post shares what we actually learned, complete with the numbers that informed our recommendations.
If you are unfamiliar with how platform engineering differs from traditional DevOps and SRE, our guide on understanding the differences between DevOps, SRE, and platform engineering provides essential context before diving into portal tooling.
Why Internal Developer Portals Matter in 2026
The internal developer portal (IDP) market has matured rapidly. Gartner forecasts that 80% of large engineering organisations will establish platform teams by 2026, up from 45% in 2022. Developer portals sit at the centre of this shift, providing a single pane of glass for service catalogues, self-service actions, scorecards, and documentation.
But here is what most vendor comparison pages omit: a portal is not a platform. A portal is one component of a broader internal developer platform. Conflating the two leads to misaligned expectations, where teams buy a portal expecting it to solve problems that require workflow orchestration, environment management, and golden paths.
With that distinction clear, let us examine the three leading options.
The Three Contenders at a Glance
| Dimension | Backstage | Port | Cortex |
|---|---|---|---|
| Type | Open-source framework (CNCF) | Commercial SaaS | Commercial SaaS |
| Data model | YAML-based, static | Flexible blueprints, no-code | Opinionated, semi-rigid |
| Plugin ecosystem | 100+ community plugins | Growing integrations via Ocean | 60+ pre-built integrations |
| Self-service actions | Build your own | Built-in action hub | Composable workflows |
| Scorecards | Via plugins | Native, customisable | Native, strong enforcement |
| AI capabilities | Limited | Agentic AI platform (2025+) | AI-powered insights |
| SSO/RBAC | Enterprise add-on or DIY | Native (Enterprise tier) | SAML SSO, RBAC included |
| Compliance | Self-managed | SOC 2 Type II, ISO 27001 | SOC 2 Type II |
| Setup time | 6-12 months | 3-6 months | 4-8 weeks (basic) |
| Ongoing maintenance | 2-15 FTEs | Vendor-managed | Vendor-managed |
Backstage: The Open-Source Framework
Backstage is Spotify’s open-source developer portal framework, now a CNCF incubating project. It commands roughly 89% market share among organisations that have adopted an IDP, largely because it was first to market and carries the Spotify brand.
What Backstage Does Well
Backstage offers unmatched extensibility. Its plugin architecture means you can integrate with virtually any tool in your stack. The software catalogue provides a centralised registry of services, and TechDocs generates documentation from Markdown files alongside your code. For organisations with the engineering capacity to invest, it can become a genuinely tailored portal.
Spotify themselves reported that their time-to-tenth-pull-request metric for new developers dropped by 55% after deploying Backstage internally.
Why We Stopped Recommending Self-Hosted Backstage
Here is where our experience diverges from the marketing. After deploying self-hosted Backstage for two client engagements, we encountered the same problems that Roadie’s research and community surveys have documented:
It is a framework, not a product. You are not installing a portal. You are committing to building one. The initial setup requires TypeScript and React expertise — skills that most platform engineering teams do not have in abundance.
Maintenance is relentless. 56% of Backstage adopters cite upgrades as their biggest pain point. Breaking changes in plugin APIs mean that upgrading Backstage versions often requires refactoring your entire plugin ecosystem. One client’s platform team spent six weeks on a single major version upgrade.
Adoption stalls at roughly 10%. Internal adoption rates hover around 10% on average, frequently because teams exhaust their capacity on maintenance before delivering the features that developers actually want to use. You end up with a portal that is technically impressive but practically empty.
Data staleness is the silent killer. Backstage relies on YAML files committed alongside service code for catalogue data. In practice, these files go stale within weeks. Developers do not update them, and without automated ingestion, your “single source of truth” becomes a single source of lies.
Backstage TCO: The Numbers Nobody Shares
Based on Roadie’s cost analysis and validated against our own client engagements:
| Cost Component | Year 1 | Year 2+ (Annual) |
|---|---|---|
| Platform engineers (3 FTEs Year 1, 2 FTEs ongoing) | $375,000 | $250,000 |
| Cloud infrastructure (hosting, database, search) | $12,000 | $12,000 |
| Opportunity cost (6-12 month delay to value) | $375,000 | $250,000 |
| Total | $762,000 | $512,000 |
These figures assume a fully loaded cost of $125,000 per Senior Platform Engineer. For organisations in London, the San Francisco Bay Area, or similar hubs, the real number is often higher.
For 100 engineers, we estimate a three-year TCO of $375,000-$750,000 per year when you factor in the maintenance team, infrastructure, and the opportunity cost of what those engineers could have built instead.
When Backstage Still Makes Sense
We do not dismiss Backstage entirely. It remains viable when:
- You have 500+ engineers and a dedicated platform team of 5+ people
- Your organisation has deep TypeScript/React expertise
- You need custom integrations that no commercial vendor supports
- You are willing to treat the portal as a product with its own roadmap
For everyone else, the managed Backstage route through Roadie or a commercial alternative is the pragmatic choice.
Port: The Flexible Builder’s Choice
Port has emerged as the most flexible commercial alternative to Backstage. Its blueprint-based data model lets you define any entity type and relationship, effectively giving you a no-code way to model your entire engineering organisation.
What Sets Port Apart
Blueprints and the context lake. Port’s data model is genuinely flexible. Unlike Backstage’s static YAML or Cortex’s semi-rigid entities, Port lets you define custom blueprints for services, environments, cloud resources, teams, or anything else. Relationships between blueprints create a knowledge graph that maps your engineering reality.
Self-service action hub. Port’s self-service actions let developers trigger workflows — spinning up environments, provisioning infrastructure, running day-2 operations — directly from the portal. This moves the portal from a read-only catalogue to an operational tool.
Agentic AI direction. Following their $100M Series C in December 2025 at an $800M valuation, Port is building an agentic engineering platform (AEP). The vision is AI agents that can resolve tickets, self-heal incidents, fix vulnerabilities, and maintain best practices — all operating within the context of your portal’s knowledge graph.
Port Pricing
| Plan | Cost | What You Get |
|---|---|---|
| Free | $0 (up to 15 users) | 10K entities, 500 automation runs, basic integrations |
| Standard | $40/seat/month (billed annually) | Up to 200 developers, full automation, SSO |
| Enterprise | Platform fee + per-seat pricing | Unlimited scale, advanced RBAC, ISO 27001, SOC 2 Type II, dedicated support |
At $40/seat/month for the Standard tier, 100 engineers would cost approximately $48,000 per year. The Enterprise tier typically runs $60-78/seat/month depending on negotiation, putting 100 engineers at $72,000-$93,600 annually.
Port’s Weaknesses
Learning curve. The flexibility that makes Port powerful also makes it complex. Building your data model, configuring blueprints, and setting up integrations requires genuine platform engineering effort. Expect 3-6 months for a full implementation.
Blueprint maintenance. While easier than maintaining Backstage code, your blueprints and integrations still require ongoing attention. Changes to your tech stack mean changes to your portal model.
Relatively young ecosystem. Port’s integration library, while growing via the open-source Ocean framework, is smaller than Backstage’s 100+ plugin catalogue.
Cortex: The Opinionated Standards Enforcer
Cortex takes a deliberately opinionated approach. Where Port gives you a blank canvas, Cortex gives you a structured framework focused on service ownership, scorecards, and engineering standards enforcement.
What Sets Cortex Apart
Scorecards as a first-class concept. Cortex’s scorecards are its defining feature. You define standards — production readiness, security compliance, documentation coverage, DORA metrics — and Cortex continuously measures every service against them. This creates accountability and visibility that other portals struggle to match.
Fastest time to value. Because Cortex is opinionated about its data model, initial setup is faster. With 60+ pre-built integrations, you can have a functional service catalogue with scorecards within weeks rather than months.
Initiatives and campaigns. Cortex lets you create targeted campaigns to drive improvements — “all services must have runbooks by Q2” — and tracks progress across the organisation. This turns the portal from a passive catalogue into an active standards enforcement tool.
Cortex Pricing
| Plan | Estimated Cost | What You Get |
|---|---|---|
| Accelerate | ~$65/user/month | Software catalogue, scorecards, 200+ workflow steps |
| Full IDP | ~$69/user/month | Everything in Accelerate plus advanced workflows, self-service actions |
| Site License | Custom pricing | Unlimited users, RBAC, SSO, dedicated support |
For 100 engineers on the Full IDP plan, expect approximately $78,000-$83,000 per year.
Cortex’s Weaknesses
Semi-rigid data model. Cortex’s opinionated structure is both its strength and limitation. If your engineering topology does not fit Cortex’s entity model, you will fight the system rather than extending it. Customisation options exist but are more constrained than Port’s blueprints.
Proprietary lock-in. Migrating away from Cortex means re-exporting your entire service catalogue, scorecards, and workflows. The more deeply you invest, the harder it becomes to leave.
Weaker self-service capabilities. While Cortex offers composable workflows, its self-service action capabilities are less flexible than Port’s action hub. If your primary use case is developer self-service rather than standards enforcement, Cortex may feel limiting.
Head-to-Head: TCO Comparison for 100 Engineers
This is the comparison table we wish we had found when we started evaluating these tools.
| Cost Factor | Self-Hosted Backstage | Port (Standard) | Cortex (Full IDP) |
|---|---|---|---|
| Software licence | $0 (open-source) | $48,000/year | $82,800/year |
| Platform engineers | $250,000-$500,000/year (2-4 FTEs) | $0 (vendor-managed) | $0 (vendor-managed) |
| Infrastructure | $12,000-$25,000/year | $0 (SaaS) | $0 (SaaS) |
| Implementation | $375,000 (Year 1 only) | $25,000-$50,000 (consulting) | $15,000-$30,000 (consulting) |
| Ongoing maintenance | $125,000-$250,000/year | Minimal (internal config) | Minimal (internal config) |
| Annual TCO (Year 2+) | $375,000-$750,000 | $48,000-$93,000 | $78,000-$83,000 |
The maths is stark. Self-hosted Backstage costs 4-10x more than commercial alternatives when you account for the engineering time required. The “free” open-source option is the most expensive choice for most organisations.
The Security and Compliance Angle
For clients in regulated industries — healthcare, financial services, government — security and compliance capabilities are often the deciding factor.
| Security Feature | Backstage | Port | Cortex |
|---|---|---|---|
| SSO/SAML | Enterprise add-on or custom | Native (Enterprise) | Native |
| Granular RBAC | Custom implementation | Native, fine-grained | Native |
| SOC 2 Type II | Self-managed | Certified | Certified |
| ISO 27001 | Self-managed | Certified (Enterprise) | Not listed |
| Audit logging | Custom implementation | Built-in | Built-in |
| Data residency | Self-hosted (you control) | Cloud-hosted (regions available) | Cloud-hosted |
Self-hosted Backstage gives you complete control over data residency, which matters for organisations with strict data sovereignty requirements. However, you bear the full burden of security patching, vulnerability management across hundreds of Node.js dependencies, and compliance auditing.
Port and Cortex handle the security operations for you, but you are trusting a third party with your engineering metadata. For most organisations, this trade-off is acceptable. For defence contractors or highly regulated entities, self-hosted Backstage (or managed Backstage via Roadie) may be the only option.
The AI and Agentic Future
The developer portal market is shifting from passive catalogues to active, AI-driven platforms. This is worth considering as you make a decision that will last 3-5 years.
Port is the clear leader here. Their agentic engineering platform vision — backed by $100M in funding — positions the portal as a command-and-control centre for AI agents that can autonomously resolve incidents, remediate vulnerabilities, and enforce standards. If AI-driven engineering operations are on your roadmap, Port’s direction is compelling.
Cortex describes itself as “AI-powered” and uses intelligence to surface insights and recommendations, but has not articulated an agentic vision as aggressively as Port.
Backstage has limited AI capabilities natively. Some community plugins integrate with LLMs, but there is no coordinated AI strategy from the CNCF project itself.
Our Decision Framework: Which Portal for Which Team
After evaluating all three for organisations of varying sizes, we developed this framework.
Under 30 Engineers
Recommendation: No dedicated portal yet.
At this size, the overhead of any portal — even Port’s free tier — outweighs the benefits. Use a well-maintained service catalogue in your wiki or a simple spreadsheet. Focus your platform engineering investment on DevOps automation and CI/CD pipelines instead.
30-100 Engineers
Recommendation: Port (Free tier or Standard).
Port’s free tier supports 15 users, which covers most platform teams. As you grow, the Standard plan at $40/seat/month offers the best value at this scale. The flexible data model lets you start simple and expand as your needs evolve.
100-200 Engineers
Recommendation: Port (Standard/Enterprise) or Cortex (Full IDP).
At this scale, the choice depends on your primary need:
- Standards enforcement and engineering metrics — choose Cortex. Its scorecards and initiatives are purpose-built for driving organisational standards.
- Flexible self-service and custom workflows — choose Port. Its blueprints and action hub shine when you need to model complex engineering topologies.
200+ Engineers
Recommendation: Port Enterprise, Cortex Enterprise, or Managed Backstage (Roadie).
At enterprise scale, all three become viable because you can justify the investment. The decision shifts to strategic alignment:
- Port if you want maximum flexibility and are betting on the agentic AI future
- Cortex if scorecards, standards enforcement, and fast time-to-value are paramount
- Managed Backstage if you need deep customisation without the maintenance burden, or have strict data residency requirements
Why We Dropped Self-Hosted Backstage
We want to be transparent about our journey. We initially championed self-hosted Backstage for clients, drawn by the CNCF backing, community ecosystem, and the appeal of open source.
After two production deployments, we changed our recommendation. Here is what happened:
Client A (150 engineers, fintech): Deployed Backstage with 2 dedicated platform engineers. After 8 months of build and customisation, adoption reached 12%. The platform team spent 60% of their time on Backstage maintenance rather than building golden paths. We migrated them to Port in Q3 2025.
Client B (300 engineers, healthcare): Attempted a more ambitious Backstage deployment with 4 FTEs. Built custom plugins for their compliance workflows. After a major Backstage version upgrade broke three plugins, the team spent 6 weeks on remediation. They explored Cortex but ultimately chose managed Backstage through Roadie to preserve their plugin investment.
The pattern was consistent: teams building Backstage became Backstage maintenance teams rather than platform engineering teams. The portal consumed the capacity meant for platform capabilities.
As Roadie’s analysis notes, the “DIY is dead” movement reflects a genuine industry shift. The engineering hours poured into portal maintenance are hours not spent on the cloud-native capabilities that actually differentiate your developer experience.
Practical Migration Tips
If you are currently on self-hosted Backstage and considering a move, here is what we have learned from actual migrations:
-
Export your service catalogue first. Document every entity, relationship, and metadata field. This becomes your requirements specification for the new platform.
-
Map your plugins to native features. List every Backstage plugin you use and identify the equivalent capability in Port or Cortex. You will find that 80% of what you built custom plugins for is available out of the box.
-
Start with the catalogue, then add self-service. Resist the temptation to migrate everything at once. Get your service catalogue accurate and adopted first, then layer on scorecards and self-service actions.
-
Plan for 8-12 weeks of parallel running. Keep Backstage available during migration so teams have a fallback. Sunset it only after adoption of the new portal exceeds your Backstage adoption rate.
-
Treat it as a platform reset. Migration is an opportunity to clean up stale data, redefine ownership, and establish the scorecards and standards you wished you had implemented from the start.
The Verdict
There is no universally correct answer, but the data points in a clear direction:
-
Self-hosted Backstage is the most expensive and highest-maintenance option. It is only justified for very large organisations (500+ engineers) with dedicated platform teams and specific customisation needs that no commercial vendor can meet.
-
Port offers the best balance of flexibility and managed operations. Its blueprint model and agentic AI direction make it the strongest long-term bet for organisations that want a portal that grows with them.
-
Cortex wins on time-to-value and standards enforcement. If your primary challenge is inconsistent service quality and you need scorecards yesterday, Cortex gets you there fastest.
For most of our clients — engineering organisations between 50 and 300 people — we now recommend starting with Port’s free tier, validating the data model against their tech stack, and upgrading to Standard or Enterprise as adoption grows. For clients whose primary pain is service quality and production readiness, we recommend Cortex.
The era of spending six months and half a million pounds building a portal from scratch is over. Whether you choose to build your DevOps capabilities in-house or work with a consulting partner, the portal should accelerate your platform — not become the platform itself.
Build Your Internal Developer Platform the Right Way
Choosing a developer portal is just one piece of the platform engineering puzzle. The real value comes from the golden paths, self-service capabilities, and operational workflows that sit on top of the portal.
Our team provides comprehensive platform engineering services to help you:
- Evaluate and implement the right portal for your organisation’s size, maturity, and technology stack
- Design golden paths that accelerate developer onboarding and reduce cognitive load
- Build self-service workflows for infrastructure provisioning, environment management, and day-2 operations
- Establish engineering standards with scorecards and initiatives that drive measurable improvement
We have deployed developer portals for organisations across fintech, healthcare, and SaaS — and we know the difference between a portal that collects dust and one that developers actually use.
Talk to our platform engineering team about your developer portal strategy