Internal Developer Platforms (IDPs)

When your platform / SRE team is the bottleneck for every "I need a new service / DB / queue / S3 bucket / dashboard" request, an IDP is the answer. Devs self-serve through a portal, the platform team controls what's possible.

The candidates

• ★ Backstage (Spotify / CNCF) — open-source IDP framework; software catalog, scaffolder, plugins. The default; large community.
• ★ Port — modern IDP-as-a-product; SaaS; clean UI; great DX. Generous free tier.
• Cortex — competitor to Port; service catalog focused.
• Roadie — Backstage-as-a-service; managed Backstage hosting.
• Humanitec — IDP focused on dynamic config + workload templates.
• OpsLevel — service catalog + scorecards.
• Configure8 / atomicwork / OpsLevel — competitive IDP/service-catalog plays.

What an IDP typically includes

• Service catalog — what services exist, who owns them, what they depend on.
• Software templates / scaffolders — Create a new service button → repo + IaC + pipelines pre-baked.
• Self-service infra — request a DB / queue / namespace; provisioned via IaC.
• Scorecards — track service health (test coverage, SLO compliance, deprecated dep, etc.).
• Single pane of glass — links to dashboards / runbooks / docs / oncall.
• Tech radar — what's recommended / experimental / deprecated.

The "platform engineering" stack underneath

Most IDPs sit on top of:

• ★ Crossplane — provision cloud infra from k8s CRDs; pairs naturally with IDPs.
• ★ GitOps (Argo CD / Flux) — for actual deployments.
• ★ Terraform / OpenTofu / Pulumi — IaC primitives.
• Backstage scaffolder + GitHub PR — the canonical "scaffold → PR → merge → deployed" loop.
• OpenTofu modules / Helm charts as the building blocks the IDP exposes.

When you need an IDP

• Multiple teams shipping services and the platform team is overloaded with "create a thing for me" tickets.
• New-team onboarding takes days because of tribal-knowledge ops.
• You can't list every service running in production.
• Scorecards / golden-paths matter for compliance.

When you don't

• Solo founder / small team — the overhead is enormous; you don't have the scale.
• Few enough services that a wiki page works.
• Your dev team is happy enough writing PRs against a deploy repo.

Backstage specifically

• Plugin ecosystem is large; integrations for most things you use.
• Operating it is real work — keeping plugins up to date, customizing, hosting.
• Roadie / Spotify Portal — managed Backstage if you don't want the ops.

Patterns to adopt

• ★ Software catalog first. Even before scaffolders. Knowing what exists is half the value.
• Scorecards drive behavior. Make "service has runbook" / "has SLO" / "has owner" measurable.
• Golden paths > rigid policy. Make the right thing easy; allow deviation with documentation.
• Plug into existing systems. Don't replace your incident tool; link to it from the portal.
• Version templates in git. Templates evolve; track who used what version.

Pick this if…

• Default OSS IDP: Backstage.
• Hosted, want minimal ops: Roadie.
• SaaS-first with great DX: Port.
• Service catalog + scorecards focus: Cortex or OpsLevel.
• You're not sure you need an IDP: you probably don't yet. Wait until the platform team is the bottleneck.