Deepak Gupta

Developer Tools · Platform Engineering

Top 6 Internal Developer Portals for 2026: Backstage vs Port vs Cortex vs OpsLevel vs Spotify Portal vs Humanitec

Build on Backstage or buy a managed portal, priced against the one thing that actually decides the outcome: whether your catalogue data is true.

By Deepak Gupta·Sep 18, 2026·16 min·6 tools compared

Platform EngineeringInternal Developer PlatformBackstageDeveloper ExperienceDeveloper ToolsDevOps

Pick Port if you want to find out whether a developer portal is worth funding. Its free tier (15 seats, 10,000 entities, no expiry) is the only one here you can run against your own estate without a sales call. Pick Backstage if you have a funded platform team of two or more engineers who will still be there next year. Pick Cortex or OpsLevel if the deliverable is a measurable service maturity number rather than a directory. Pick Humanitec only if the real bottleneck is that developers cannot get an environment, because it is a platform orchestrator with a portal, not a portal.

The build-versus-buy question is usually framed as a licence cost question, and that framing is wrong. Backstage's licence cost is zero. Its operating cost is one to three platform engineers permanently assigned, and most of their time goes to plugin compatibility and upgrades rather than anything your developers will notice. A managed portal at $30 to $40 per seat per month runs roughly $36,000 to $48,000 a year for 100 engineers, which is less than one senior platform engineer in most markets. Framed that way, the decision resolves quickly in most organisations.

Note

The honest weakness of this entire category: internal developer platform projects almost never fail on tooling. They fail because the catalogue data is not true. Ownership fields point at people who left, entries describe decommissioned systems, runbook links 404, and within two quarters nobody believes the portal. Once trust is gone, no feature recovers it. Decide who is accountable for catalogue accuracy before you decide which product to buy.

That is also why the evaluation criterion that matters most is how a portal keeps its data honest. Checks that inspect the repository directly, as OpsLevel's repo checks and package version checks do, correspond to something real. Checks that read a declared catalog-info.yaml describe intent, and intent rots. Weight automated ingestion and code-reading checks far above the feature matrix.

Two market facts changed the shortlist this year and are worth stating explicitly. Backstage is still at CNCF Incubating maturity, not Graduated, despite very large adoption. And managed Backstage is a real purchasable category rather than a false choice. Spotify sells Portal for Backstage, and Roadie has sold managed Backstage since 2020. Roadie's pricing page labels its tiers for existing subscribers as of September 2026, so confirm availability before shortlisting it.

Every price on this page was read from the vendor's own pricing page in September 2026. Where a vendor publishes no price, this page says so rather than repeating a number from a comparison site. If your next question is about the CI and deployment layer underneath the portal, see the CI/CD platform comparison and the Kubernetes management platform comparison.

Quick Comparison

Portal	Build or Buy	Free Tier	Published Price	Where It Breaks
Backstage (CNCF)	Build: a framework, not a product	Apache 2.0, unlimited, self-hosted	$0 licence, one to three engineers of standing cost	Plugin and upgrade maintenance never ends
Port	Buy: managed, no Backstage code	Up to 15 seats and 10,000 entities	$30/seat/mo Basic, $40/seat/mo Standard	Flexible data model means you own the modelling mistakes
Cortex	Buy: managed, scorecard-first	No free tier published	Not published, per-seat quote only	No self-serve price, so no cheap pilot
OpsLevel	Buy: managed, maturity-check-first	No free tier published	Not published, per-developer quote only	Standard plan caps at 50 users
Spotify Portal for Backstage	Buy: managed Backstage from its creator	No free tier published	Not published, contact sales	You inherit Backstage's model without Backstage's community support terms
Humanitec	Buy: orchestrator first, portal second	No free tier, trial only	$2,199/mo Teams (5 users), $5,499/mo Pro (50 users)	Highest floor price here and it is not primarily a portal

Backstage (CNCF)

Build or Buy: Build: a framework, not a product
Free Tier: Apache 2.0, unlimited, self-hosted
Published Price: $0 licence, one to three engineers of standing cost
Where It Breaks: Plugin and upgrade maintenance never ends

Port

Build or Buy: Buy: managed, no Backstage code
Free Tier: Up to 15 seats and 10,000 entities
Published Price: $30/seat/mo Basic, $40/seat/mo Standard
Where It Breaks: Flexible data model means you own the modelling mistakes

Cortex

Build or Buy: Buy: managed, scorecard-first
Free Tier: No free tier published
Published Price: Not published, per-seat quote only
Where It Breaks: No self-serve price, so no cheap pilot

OpsLevel

Build or Buy: Buy: managed, maturity-check-first
Free Tier: No free tier published
Published Price: Not published, per-developer quote only
Where It Breaks: Standard plan caps at 50 users

Spotify Portal for Backstage

Build or Buy: Buy: managed Backstage from its creator
Free Tier: No free tier published
Published Price: Not published, contact sales
Where It Breaks: You inherit Backstage's model without Backstage's community support terms

Humanitec

Build or Buy: Buy: orchestrator first, portal second
Free Tier: No free tier, trial only
Published Price: $2,199/mo Teams (5 users), $5,499/mo Pro (50 users)
Where It Breaks: Highest floor price here and it is not primarily a portal

1

Backstage

Best Open Source

Best for: Organisations with a funded platform team of at least two or three engineers who need a portal shaped exactly to their internal model and can carry the upgrade work forever

“Backstage is the default answer and the most commonly regretted one. It is a CNCF project created at Spotify, Apache 2.0, with an enormous plugin ecosystem, and it will model almost anything you want. It is also a framework rather than a product: you write TypeScript, you run the service, and you own every plugin upgrade. Teams that succeed with it staffed it properly. Teams that failed treated it as an install.”

Pros

Apache 2.0 with no seat count, entity count, or feature gate, so cost does not scale with headcount
CNCF project since September 2020 and at Incubating maturity since March 2022, with governance that does not depend on one vendor's roadmap
The largest plugin ecosystem of any portal, covering CI, cloud, cost, security, and documentation surfaces
The catalogue model (Component, System, API, Resource, Group) is a genuinely good ontology that the commercial products largely copied
Software Templates give you golden paths that create a repo, wire CI, and register the new service in one action

Cons

It ships as a scaffolded app you must build and maintain, not an installable product, and the reference deployment is your problem
Plugin compatibility across Backstage versions is the recurring tax: an upgrade can break several plugins at once and there is no vendor to escalate to
Authentication, RBAC, and multi-tenant concerns need real work before it is safe to expose to the whole engineering org
There is no built-in answer to stale catalogue data, so an unmaintained catalog-info.yaml quietly rots and takes trust with it
Still Incubating rather than Graduated in CNCF, which some procurement teams treat as a risk flag even though adoption is enormous

Honest Weakness: Backstage's licence cost is zero and its real cost is a standing engineering commitment. Budget one to three engineers continuously, not a quarter of project time, and understand that most of that work is upgrade and plugin maintenance rather than features your developers will notice. If you cannot name the people who will own it next year, buying a managed portal is the cheaper decision even at $40 a seat.

Framework, Not Product

The single most expensive misunderstanding about Backstage is that it is something you install. It is a scaffolding CLI that produces a TypeScript monorepo you then own. Adding a plugin means editing application code and redeploying. That is a deliberate design choice and it is why Backstage can model an organisation's real topology better than a closed product can, but it means the comparison against Port or Cortex is not a comparison of two products. It is a comparison between a product and a project.

The Upgrade Tax Is The Line Item

Backstage releases move quickly and plugins are maintained by many different parties at different speeds. The predictable failure is an upgrade that breaks three plugins, none of which is your code, none of which has a support contract behind it. Teams that run Backstage well pin versions deliberately, keep the plugin count small on purpose, and schedule upgrade work as recurring capacity. Teams that discover this pattern eighteen months in are the ones who end up buying a managed portal anyway.

Free. Apache 2.0 licence, self-hosted, no seat or entity limits. Real cost is the platform engineers who run it, plus compute and a Postgres instance.

Visit Backstage

2

Port

Best Overall

Best for: Teams that want a working portal in weeks rather than quarters, with a free tier large enough that a pilot costs nothing and a data model flexible enough to describe infrastructure Backstage's ontology does not cover

“Port is the strongest buy-side option for most organisations because it is the only one on this list with a free tier you can genuinely run a pilot on: 15 seats and 10,000 entities, no credit card. It is not built on Backstage, which cuts both ways. You get no plugin ecosystem inheritance, but you also get no Backstage upgrade tax. Its data model is a blank schema rather than a fixed ontology, which is powerful and means the modelling mistakes are yours.”

Pros

Free plan covers up to 15 active users and 10,000 entities with the full platform and no time limit, so a real pilot costs nothing
Published self-serve pricing at $30/seat/month (Basic, up to 50 seats, 50,000 entities) and $40/seat/month (Standard, up to 200 seats, 250,000 entities)
Port states every tier gets the full platform, with tiers differing on capacity, workspaces, and support rather than gating features
Fully custom data model: you define blueprints and relations rather than fitting your world into a fixed Component and System ontology
Self-service actions and scorecards are first-class, so the portal can both show state and change it

Cons

Not Backstage-based, so the large open plugin ecosystem does not carry over and integrations are whatever Port ships or you build against its API
Enterprise pricing is not published, so the step above 200 seats requires a sales conversation with no public anchor
A fully flexible schema means there is no opinionated starting ontology, and a badly modelled catalogue is harder to fix later than a rigid one
Workspaces (5 on Standard, 20 on Enterprise) are a real constraint for organisations that want hard separation between business units

Honest Weakness: Port's flexibility is a trap if nobody owns the model. Because you define blueprints from scratch, two teams will model the same concept two ways within a quarter unless someone is explicitly accountable for the schema. The portals that rot fastest are the ones where the flexible model was treated as a feature that removed the need for a decision. Port will not stop you making that mistake, and its entity caps (10,000 free, 50,000 Basic) arrive quickly once you start ingesting cloud resources rather than just services.

The Only Real Free Pilot

Every other commercial portal here requires a sales call before you can see it against your own data. Port's free tier is 15 seats and 10,000 entities with no expiry, which is enough to ingest a mid-sized estate and find out in two weeks whether your catalogue data is good enough to build on. That matters more than any feature comparison, because the question that decides an internal developer platform project is not which portal is best, it is whether your ownership data is true. Port lets you answer that for free.

Free: $0, up to 15 seats, 10,000 entities, community support. Basic: $30/seat/month, up to 50 seats, 50,000 entities. Standard: $40/seat/month, up to 200 seats, 250,000 entities, 5 workspaces. Enterprise: custom, 20 workspaces. Prices as published on Port's pricing page, September 2026.

Visit Port

3

Cortex

Best for Enterprise

Best for: Engineering organisations whose actual goal is measurable service maturity, where scorecards and production readiness standards are the deliverable and the catalogue is the means

“Cortex is the portal to shortlist when leadership is asking for a number rather than a directory. It is built around scorecards and initiatives: define what a production-ready service looks like, score everything against it, and drive the gap down. That framing is real and it is the reason Cortex wins deals against Backstage. The cost of that framing is that Cortex publishes no price at all, so the evaluation starts with a sales cycle rather than a trial.”

Pros

Scorecards and initiatives are the product, not an add-on, so service maturity reporting works out of the box rather than as a plugin you build
Strong catalogue ingestion from cloud providers, git hosts, and observability tools, which reduces the manual YAML problem that sinks Backstage rollouts
Self-service actions and templates cover the golden path use case that most buyers are actually shopping for
Per-seat pricing scales with the team, and Cortex states quotes are built to organisation size rather than fixed tiers

Cons

No published pricing of any kind and no self-serve tier, so every evaluation begins with a sales conversation
No free tier documented on the pricing page, which makes a low-commitment proof of value harder than with Port
Scorecard-first framing suits organisations with an existing standards culture, and lands badly where teams have not agreed what good looks like
Not Backstage-based, so no inheritance from the open plugin ecosystem

Honest Weakness: Cortex measures compliance with standards you have to write first. If your organisation has not agreed what a production-ready service is, Cortex will faithfully score every team against a document nobody believes in, and the portal becomes the thing engineers resent rather than use. The tool is not the hard part. The hard part is the standards conversation, and buying Cortex before having it is how scorecards turn into shelfware.

Scorecards Are A Political Instrument

A scorecard makes a previously invisible judgement public: this service is below standard, and here is who owns it. That is exactly what platform leaders want and exactly what makes rollouts fail when the standard was written by one team and applied to forty. Cortex is good at the mechanics. Treat the mechanics as the easy 20 percent, and budget the other 80 percent for agreeing the standard with the teams it will be applied to.

Not published. Cortex states pricing scales with team size and is quoted per seat on request. No free tier is listed on its pricing page as of September 2026.

Visit Cortex

4

OpsLevel

Runner Up

Best for: Mid-sized engineering organisations that want catalogue plus maturity checks plus self-service actions as a managed product, with campaigns to actually drive remediation to completion

“OpsLevel sits very close to Cortex in scope and differentiates on campaigns: a structured way to push a specific fix across every service and track it to done, rather than leaving a scorecard red indefinitely. Its Standard plan covers up to 50 users and Enterprise is unlimited. Like Cortex it publishes no price, and unlike Port it offers no free tier, so the cheapest way in is a guided tour rather than your own data.”

Pros

Campaigns turn a scorecard failure into a tracked remediation programme with owners and deadlines, which is the step most portals leave to spreadsheets
Unlimited catalogued components on the Standard plan, so the catalogue itself is not a metered resource
Repo checks and package version checks read the code rather than trusting a YAML file, which resists catalogue rot better than declarative-only approaches
SSO is included on the Standard plan rather than reserved for Enterprise, which is unusual and worth noting

Cons

No published per-user price, so budgeting requires a quote
No free tier; the evaluation path is a demo or a guided product tour rather than running it against your own estate
Standard is capped at 50 users, so organisations just past that line are pushed to Enterprise and custom pricing
Feature overlap with Cortex is high enough that the choice between them usually comes down to the sales process and the campaigns feature

Honest Weakness: OpsLevel and Cortex are close enough in capability that most buyers cannot honestly differentiate them on a feature matrix, which means the decision gets made on price and the sales relationship, neither of which is visible before you engage. If you are a 60-person engineering org, you are above the Standard plan's 50-user cap and immediately in a custom quote with no public anchor to negotiate against. That is a weaker buying position than Port's published $40 a seat.

Checks That Read Code, Not YAML

The reason catalogues rot is that they describe intent rather than reality: a file says a service has a runbook, and the link 404s. OpsLevel's repo checks and package version checks inspect the repository directly, so a passing check corresponds to something that is actually in the code. That is a meaningful defence against the single biggest failure mode in this category, and it is worth weighting heavily in a bake-off against any portal whose checks only read declared metadata.

Not published. Standard covers up to 50 users, Enterprise is unlimited users, both quoted per developer on request. No free tier listed as of September 2026; OpsLevel offers demos and self-guided product tours instead.

Visit OpsLevel

5

Spotify Portal for Backstage

Honorable Mention

Best for: Organisations that have decided on Backstage's model and want it from the team that invented it, without running the TypeScript monorepo themselves

“Spotify sells a managed Portal built on Backstage plus a plugin bundle (Insights, Soundcheck, RBAC, Skill Exchange) that covers the gaps open Backstage leaves. If your reason for choosing Backstage is the model and the ecosystem rather than the zero licence fee, this removes the part that actually costs money. Spotify publishes no pricing, and its dedicated pricing page returned a 404 when checked, so the entire commercial conversation happens with sales.”

Pros

Built on Backstage by Backstage's creators, so the catalogue model and plugin ecosystem transfer rather than being re-learned
No-code setup and managed hosting remove the scaffolded-monorepo and upgrade burden that sinks most self-hosted Backstage projects
The Spotify plugin bundle fills the real gaps in open Backstage: RBAC, quality checks via Soundcheck, and usage analytics via Insights
A credible migration path in both directions, since the underlying model is the same open project

Cons

No published pricing at all, and the pricing URL on the product site was a 404 when checked in September 2026
No free tier documented, so evaluation requires contacting Spotify
Core value depends on the paid plugin bundle, which means the comparison against free Backstage is partly a comparison against plugins you could otherwise build
Smaller commercial track record as a vendor than Port, Cortex, or OpsLevel, whose entire business is selling this product

Honest Weakness: You are buying a commercial product from a company whose main business is music streaming, on pricing that is not published anywhere. That is a governance and procurement risk independent of how good the product is, and it is the reason a lot of Backstage shops end up at Roadie or a self-hosted build instead. The technical inheritance from Backstage is genuine; the commercial certainty is the weakest of the paid options here.

Managed Backstage Is A Real Category

Spotify Portal is not the only way to buy Backstage rather than build it. Roadie has sold managed Backstage since 2020 and publishes a Teams tier at $24 per developer per month. As of September 2026 its pricing page labels its plans for existing subscribers, so confirm availability for a net-new purchase before shortlisting it. The category matters because it resolves the false choice most teams think they face: Backstage's model without Backstage's maintenance bill is a purchasable thing.

Not published. Spotify directs prospects to contact sales for Portal and the Plugins bundle. The product site's pricing path returned a 404 when checked in September 2026.

Visit Spotify Portal for Backstage

6

Humanitec

Best for Enterprise

Best for: Platform teams whose real problem is environment and infrastructure provisioning rather than discovery, and who want a Platform Orchestrator with a portal attached rather than the reverse

“Humanitec is on this list because buyers keep shortlisting it against portals, and it is important to say plainly that it is a different product. Its core is a Platform Orchestrator that generates application configuration and provisions infrastructure from a workload spec, with Portal as a frontend and Score as the CLI. If your pain is that developers cannot get an environment, Humanitec is more relevant than any portal here. If your pain is that nobody knows who owns a service, it is the wrong purchase at the highest price on this page.”

Pros

Publishes real prices, which is rare in this category and makes budgeting possible without a sales cycle
The Platform Orchestrator genuinely solves dynamic environment provisioning, which no catalogue-first portal does
Score gives a workload specification that is portable across environments and decouples developer intent from platform implementation
Self-hosted deployment is offered for regulated and private-cloud requirements

Cons

Teams tier is $2,199 per month for 5 users and a maximum of 2 projects, the highest entry price on this page by a wide margin
No free tier at all; evaluation is a trial or a paid MVP onboarding programme
RBAC is not available on Teams, and SAML and audit logs are Enterprise-only, so the cheap tier is not enterprise-shaped
Sold as a platform orchestrator, so a buyer shopping for a service catalogue will find the portal component thinner than a dedicated portal

Honest Weakness: Humanitec is priced as an infrastructure platform and shortlisted as a portal, and that mismatch is where deals go wrong. At $2,199 a month for five users you are paying orchestrator prices for a catalogue you could get free on Port, and the tier that includes RBAC starts at $5,499 a month. Buy Humanitec when the provisioning problem is the expensive one. Do not buy it because the portal screenshot looked good.

Orchestrator Versus Portal

The clean split is this. A portal answers what exists, who owns it, and is it healthy. An orchestrator answers give me a staging environment with a Postgres and an S3 bucket, correctly configured for this environment, without a ticket. Backstage, Port, Cortex, OpsLevel, and Spotify Portal are portals. Humanitec is an orchestrator with a portal. Buying the wrong one is the most common expensive mistake in platform engineering procurement, because both categories describe themselves as internal developer platforms.

Teams: $2,199/month (EUR 1,999) for 5 users, max 2 projects, up to 5 environments per project, email support. Pro: $5,499/month (EUR 4,999) for 50 users, up to 10 projects, unlimited environments, RBAC. Enterprise and Self-Hosted: custom. 10 percent discount on annual billing. No free tier; trials available without a credit card. Prices as published on Humanitec's pricing page, September 2026.

Visit Humanitec

Which One Should You Pick?

Use Case	Our Recommendation
You want to find out whether an internal developer portal is worth funding, before asking for budget	Port's free tier. 15 seats and 10,000 entities with no time limit and no credit card is enough to ingest a real estate and discover whether your ownership data is accurate. That question, not the tooling question, decides whether the project succeeds.
You have a funded platform team of two or more engineers and need a portal shaped to a model no vendor sells	Backstage. Apache 2.0, the best ontology in the category, the largest plugin ecosystem, and no per-seat cost. Budget the engineers as a permanent line item rather than a project, because plugin and upgrade maintenance never ends.
Leadership wants a measurable number for service maturity and production readiness	Cortex or OpsLevel. Both are scorecard-first managed products. Pick OpsLevel if you also need campaigns to drive a specific fix to completion across every service; pick Cortex if initiatives and maturity reporting are the primary deliverable. Neither publishes a price, so run both quotes together.
You are committed to the Backstage model but cannot staff the maintenance	Spotify Portal for Backstage, or a managed Backstage vendor such as Roadie. You keep the catalogue model and ecosystem and drop the monorepo upgrade work. Neither Spotify nor Roadie publishes self-serve pricing you can act on without a conversation, so start the commercial thread early.
Developers cannot get an environment without filing a ticket, and that is the actual bottleneck	Humanitec. This is an orchestration problem, not a discovery problem, and no catalogue-first portal solves it. Expect to start at $2,199 per month for 5 users, and note RBAC begins at the $5,499 Pro tier.
Your catalogue keeps going stale and nobody trusts it any more	Prioritise checks that read the repository rather than declared metadata, which is OpsLevel's repo checks and package version checks, or equivalent automated ingestion in Port and Cortex. No portal fixes this by itself. Assign an owner for catalogue accuracy before you assign a budget for a portal.

How we evaluated

Internal developer portals are bought on demos and judged eighteen months later on whether anyone still trusts the catalogue. This comparison weights the things that decide that outcome over the things that decide a bake-off.

Each product was assessed on the criteria in the comparison table above:

Build or buy: whether the product is a framework you own and operate, a managed portal, or an orchestrator shortlisted as a portal. The categories are not interchangeable and buying across them is the most common expensive mistake here.
Free tier: what you can actually run against your own data without a sales conversation, including seat and entity caps.
Published price: the real, vendor-published number. Four of the six vendors on this page publish nothing, and that is reported as a finding rather than filled in from a third party.
Catalogue integrity: whether checks inspect the repository or only read declared metadata, because declared metadata is what rots.
Where it breaks: the failure mode each product actually has in production, stated plainly.

What we verified, and when

Last verified: September 2026. Every price, seat cap, and entity limit on this page was read from the vendor's own pricing page or documentation in September 2026:

Port pricing for the Free, Basic, Standard, and Enterprise seat and entity limits.
Humanitec pricing for the Teams and Pro monthly prices, user and project caps, and the RBAC, SAML, and audit-log tier boundaries.
Cortex pricing and OpsLevel pricing, both of which state a per-seat model and publish no number.
CNCF's Backstage project page for the acceptance date (8 September 2020) and current Incubating maturity level (since 15 March 2022).
Spotify's Backstage product site for the Portal and Plugins product scope. Its pricing path returned a 404 when checked, which is why this page reports no Spotify price rather than estimating one.
Roadie pricing for the $24 per developer per month Teams figure and the existing-subscribers labelling noted in the Spotify entry.

Where a vendor does not publish a price, this page says so. No figure here was carried from another comparison site or inferred from an aggregator.

What we could not verify

Cortex, OpsLevel, and Spotify publish no pricing, so the cost comparison across this category is genuinely incomplete and the table says so rather than guessing. Roadie's willingness to take net-new customers could not be confirmed from its own pages.

Note

Editorial independence: this is a vendor-neutral comparison with no paid placements, sponsorships, or affiliate links. Rankings reflect fit for the stated use cases, not commercial relationships. No product on this page was tested hands-on; the evaluation rests on vendor documentation, published pricing, and CNCF project records, all cited above.

Frequently Asked Questions

Should we build on Backstage or buy a managed internal developer portal?

Answer it with headcount, not features. Backstage costs nothing to licence and realistically needs one to three engineers permanently assigned, mostly to plugin compatibility and upgrades rather than to features your developers will notice. A managed portal at $30 to $40 per seat per month costs roughly $36,000 to $48,000 a year for 100 engineers, which is less than one senior platform engineer in most markets. If you cannot name the people who will own Backstage twelve months from now, buy. If you have a funded platform team and a model no vendor sells, build.

Why do internal developer platform projects fail?

Almost never because of the tooling. They fail because the catalogue data is not true. Ownership fields point at people who left, service entries describe systems that were decommissioned, runbook links 404, and within two quarters engineers stop believing what the portal says. Once trust is gone, no feature recovers it. The controllable version of this problem is to decide who is accountable for catalogue accuracy before you pick a product, and to prefer checks that inspect the repository over checks that read a declared YAML file.

Is Backstage actually free?

The software is Apache 2.0 with no seat limit, entity limit, or feature gate, so the licence cost is genuinely zero. The operating cost is not. Backstage ships as a scaffolded TypeScript monorepo you build and run yourself, plugins are maintained by many different parties on different schedules, and an upgrade can break several at once with no vendor to escalate to. The honest budget line is continuous platform engineering capacity plus compute and a Postgres instance, not zero.

Which internal developer portal has a real free tier?

Port is the only one of these six with a usable free tier: up to 15 active users and 10,000 entities, full platform, community support via Slack, no credit card and no expiry. Backstage is free in the open-source sense, which is a different thing. Cortex, OpsLevel, Spotify Portal, and Humanitec publish no free tier as of September 2026; Humanitec offers trials without a credit card, and OpsLevel offers demos and self-guided tours.

What is the difference between an internal developer portal and a platform orchestrator?

A portal answers questions about what exists: which services run, who owns them, what their maturity score is, and how to create a new one from a template. An orchestrator changes infrastructure: it takes a workload specification and provisions a correctly configured environment without a ticket. Backstage, Port, Cortex, OpsLevel, and Spotify Portal are portals. Humanitec is an orchestrator that also ships a portal. Both categories market themselves as internal developer platforms, which is why buyers routinely shortlist products that do not solve their problem.

Is Backstage a graduated CNCF project?

No. Backstage was accepted into the CNCF on 8 September 2020 and moved to Incubating maturity on 15 March 2022, and it remains at Incubating as of September 2026. That is worth knowing if your procurement process treats CNCF graduation as a maturity signal. Adoption is very large regardless: the CNCF and Spotify report thousands of adopting organisations, and Backstage ranks among the highest CNCF projects by development velocity.

How many entities will we actually have in the catalogue?

More than you expect, and it matters because Port meters on entities: 10,000 on Free, 50,000 on Basic, 250,000 on Standard. If the catalogue holds only services and teams, a 500-engineer org sits in the low thousands. If you ingest cloud resources, Kubernetes workloads, databases, and queues, the same org crosses 50,000 quickly. Decide what belongs in the catalogue before you pick a tier, because ingesting everything is the fastest route both to a bigger bill and to a catalogue nobody can navigate.

About the author

Deepak Gupta is the founder and creator of LoginRadius, a customer identity platform he built and scaled to over a billion users. He is now the founder of GrackerAI, a GEO platform for B2B SaaS and cybersecurity teams, and has spent more than 15 years building identity and security products.

Related Comparisons

Incident Response

Top 6 Incident Management and On-Call Platforms for 2026: incident.io vs PagerDuty vs Rootly vs Grafana IRM vs Better Stack vs FireHydrant

6 tools compared

Workflow Orchestration

Top 6 Workflow Orchestration Platforms for 2026: Temporal vs Airflow vs Prefect vs Dagster vs Kestra vs Windmill

6 tools compared

Container Infrastructure

Top 8 Container Registries for 2026: Pull Rate Limits, Retention, Scanning, and Egress Priced

8 tools compared

Backend Platform

Top 7 Supabase Alternatives for Startups (2026)

7 tools compared