Backstage holds roughly 89% share among portal frameworks, yet adoption inside the companies that run it averages around 10% of developers. Market share and actual usage are two very different numbers, and picking an IDP on the first one is how teams end up with a portal nobody opens.
"Internal developer platform" now covers at least three kinds of product. Some are developer portals: a catalog of services, owners, docs and scorecards. Some are platform orchestrators: engines that turn a developer request into real infrastructure. And some try to do both. Comparing a portal to an orchestrator on a feature grid is like comparing a map to a car, so this list sorts the options by what they do before ranking anything.
Eight tools are covered below. Every entry, Cycloid included, gets a "where it fits" and a "watch out for" note.
How This List Was Built
Four questions decided how each tool is described:
- Does it only show what exists, or can it also create it? Catalog-only and provisioning tools solve different problems.
- What does it cost to run? That covers licence price and the engineers needed to keep it alive.
- Where can it be deployed? SaaS only, self-hosted, or both matters for regulated teams.
- How long until developers get value? Weeks and multi-quarter builds lead to very different outcomes.
Pricing changes often. Treat every figure as a starting point and confirm on the vendor's page before budgeting.
Best Internal Developer Platforms Comparison
| Platform | Type | Provisions infrastructure? | Deployment | Pricing model |
|---|---|---|---|---|
| Cycloid | Portal plus orchestrator | Yes, Terraform-based | SaaS, self-hosted, dedicated SaaS | Free up to 5 users, then per-user |
| Backstage | Portal framework (open source) | Only through plugins you build | Self-hosted | Free licence, high engineering cost |
| Port | Portal, no-code | Via self-service actions | SaaS | Free tier, then per-seat tiers |
| Humanitec | Platform orchestrator | Yes | SaaS, self-hosted option | Custom |
| Cortex | Portal, scorecard-led | Limited | SaaS | Custom |
| OpsLevel | Portal, maturity-led | Limited | SaaS | Per-user tiers |
| Roadie | Managed Backstage | Via plugins | SaaS | Per-seat |
| Compass (Atlassian) | Portal inside Atlassian | Limited | SaaS | Per-user, tied to Atlassian plans |
Top Internal Developer Platforms in 2026
1. Cycloid!
Cycloid describes itself as an internal developer portal and platform in one: a portal you customise, Blueprints and policy, to deploy, run, promote and compose. The practical result is that a developer picks an environment from a form, and Terraform runs behind it under pre-approved policy. StackForms handle the forms, and Stacks package the approved templates. Around that sit pre-deployment cost estimation, cost management, carbon footprint tracking, asset inventory, InfraView, and an MCP server for governed AI assistance.
Deployment is the broadest on this list: SaaS, self-hosted, or dedicated SaaS. SaaS customers host their own Concourse workers so pipeline execution stays in their environment. The European Commission runs it at a scale of up to 5,000 developers, according to the Cycloid site.
Pricing: Free for up to 5 users on SaaS. Advanced is listed at \$45 per user per month billed annually, roughly \$27,000 a year at 50 named users. Enterprise is custom, with a listed starting price of \$99, per Cycloid's pricing page.
Where it fits: Teams whose real bottleneck is infrastructure provisioning, especially Terraform-heavy shops, multi-cloud estates, and regulated organisations that need self-hosted or sovereign deployment.
Watch out for: The plugin ecosystem is smaller than Backstage's, which Cycloid itself concedes. Teams that run purely Kubernetes-native workloads without infrastructure as code may find the Terraform focus more than they need.
2. Backstage!
Backstage is the open-source framework Spotify released, now a CNCF project, and the reason the category exists. It gives you a software catalog, TechDocs, scaffolding templates and a plugin system. You assemble the platform yourself.
Where it fits: Organisations with a dedicated platform team, strong TypeScript skills, and a wish to own every layer.
Watch out for: Self-hosted instances commonly need 3 to 15 engineers to maintain, and teams often report spending 30 to 40% of platform time on plugin upkeep. It is a framework, not a finished product. Out of the box it lists things; it does not provision them.
3. Port!
Port is a no-code developer portal built around a flexible data model. Teams define their own blueprints for services and environments, then attach scorecards and self-service actions.
Where it fits: Teams that want a catalog live in weeks without writing a React app, and that model their organization in unusual ways.
Watch out for: Port is strongest as the front door. The actual provisioning still runs through the pipelines you connect behind it.
4. Humanitec!
Humanitec built the Platform Orchestrator category and is tied to the Score workload specification. Developers describe what a workload needs, and the orchestrator resolves that into the right resources per environment.
Where it fits: Organisations that want dynamic configuration management, where one workload definition deploys differently to dev, staging and production.
Watch out for: It is an orchestrator, not a portal, so many teams pair it with a catalog. That is extra integration work if you want one product.
5. Cortex!
Cortex focuses on service quality. Its scorecards and initiatives let engineering leaders define standards and track which services meet them.
Where it fits: Larger engineering organisations where the main problem is standards drift and ownership gaps.
Watch out for: If developers mainly complain about waiting for environments, a scorecard will describe the problem without solving it.
6. OpsLevel!
OpsLevel is a catalog and maturity tool with a reputation for quick onboarding. It maps services, owners and dependencies, and grades them against rubrics.
Where it fits: Teams that want a clean service catalog without much setup, and handle provisioning elsewhere.
Watch out for: It leans toward visibility over automation. Check what self-service actions your use case needs.
7. Roadie
Roadie is hosted Backstage. It runs the framework for you, patches it, and curates a set of plugins.
Where it fits: Teams that have decided on Backstage but do not want to staff a team to run it.
Watch out for: You inherit Backstage's design, including the fact that provisioning depends on plugins and templates you configure.
8. Compass (Atlassian)
Compass is Atlassian's component catalog, integrated with Jira, Bitbucket and Confluence.
Where it fits: Companies already standardised on Atlassian, where shared identity and data remove friction.
Watch out for: It is lighter on provisioning and orchestration than dedicated platforms, and its value drops outside the Atlassian ecosystem.
Best-Practice Guidelines for Choosing and Rolling Out an IDP
These guidelines come from platform engineering research and DORA findings, and they apply whichever tool you pick.
- Start with one workflow, not the whole platform. Platforms that try to solve everything at launch fail far more often than those that begin with a single high-pain workflow such as environment provisioning.
- Measure before you launch. Capture lead time, provisioning time, deployment frequency and ticket volume first. Roughly 30% of platform teams do not measure success at all, so many cannot show value at the first budget review.
- Give the platform an owner. Teams with a dedicated platform product manager report noticeably higher internal adoption than engineer-only teams.
- Hide governance in the defaults. Policy-as-code at request time removes approval queues instead of adding one.
- Give developers clear feedback on outcomes. DORA's platform engineering research links clear feedback after a task is submitted to a better developer experience.
- Run a two-week proof of concept on the workflow developers hit most often, and judge tools on that instead of on feature grids.
How to Choose
- Developers wait days for infrastructure: Cycloid or Humanitec first.
- Nobody knows who owns what: a catalog such as Port, OpsLevel, Cortex or Compass.
- You want total control and have the headcount: Backstage, or Roadie if you would rather not run it.
- You are regulated or need EU sovereignty: check deployment options first. Cycloid's self-hosted and dedicated SaaS routes narrow the list quickly.
- Everyone lives in Atlassian: Compass deserves a look for integration reasons.
Frequently Asked Questions
What is the difference between an internal developer platform and a developer portal? A portal is the interface: a catalog, docs and scorecards. An IDP is the whole capability, including the automation that provisions and deploys. Many products called IDPs are portals with the automation left to you.
Is Backstage an internal developer platform? It is a framework for building one. Backstage supplies the catalog and plugin system, but you build the provisioning and workflows on top.
How much does it cost to run Backstage? The licence is free, but self-hosted instances commonly require 3 to 15 engineers, and plugin maintenance can take 30 to 40% of platform team time.
Which IDP is best for a small team? Tools with free tiers and fast setup, such as Cycloid (free up to 5 users on SaaS) or Port, let you validate the approach before paying.
Can I combine tools? Yes. A common pattern is a portal in front and an orchestrator behind it. The trade-off is integration work you own.
How long does an IDP rollout take? Portal-first tools can show a working catalog in weeks. Custom Backstage builds often run for multiple quarters before broad adoption.
