Internal Developer Platforms (IDPs): Backstage, Port, Cortex, OpsLevel, Roadie, Humanitec, Mia-Platform

⬅️ DevOps & Tools Overview

If you're a B2B SaaS at $20M+ ARR with 30+ engineers, you've started feeling the pain — services proliferate, ownership is unclear, deploy paths fragment, new engineers take weeks to ship their first PR. The naive answer: more docs in Notion. The structured answer: an Internal Developer Platform (IDP) — Backstage, Port, Cortex, OpsLevel — that catalogs services + ownership, scaffolds new ones, embeds standards, surfaces dashboards, runs self-serve actions. IDPs scale engineering productivity at the layer where Slack-and-docs stops working. The right pick depends on whether you want OSS (Backstage) vs managed (Port / Cortex / OpsLevel) and what depth of platform you need.

TL;DR Decision Matrix

The first decision is OSS vs managed: Backstage is OSS but requires significant engineering investment; managed IDPs (Port / Cortex / OpsLevel) trade $$ for time-to-value. The second decision is platform depth: service catalog + scorecards (Cortex / OpsLevel) vs full platform (Backstage / Port / Humanitec).

Decide What You Need First

Service catalog + ownership (the 50% case)

Engineers can't find which team owns what; oncall pages wrong people; new hires lost.

Right tools:

• Backstage — OSS catalog
• Port — managed
• Cortex OR OpsLevel — catalog + scorecards
• Roadie — hosted Backstage

Self-serve scaffolding (the 25% case)

Engineers need to create new services / repos / pipelines without DevOps tickets.

Right tools:

• Backstage with templates
• Port with self-service actions
• Humanitec for deployment self-service

Service maturity / production-readiness (the 15% case)

Define standards (testing, monitoring, on-call); track which services meet them.

Right tools:

• Cortex — scorecard leader
• OpsLevel — alternative
• Backstage scaffolding plugin

Cloud dev environments (the 10% case)

Engineers run dev locally with hour-long setup; CDEs solve.

Right tools:

• Gitpod — CDE leader
• GitHub Codespaces — GitHub-native
• Coder — self-hosted CDE
• (Distinct from IDPs; can complement)

Provider Deep-Dives

Backstage — OSS leader

Open-sourced by Spotify 2020. Massive ecosystem.

Pricing: free (self-host); ~3-12 months engineering investment to build internal version.

Features: software catalog, software templates (scaffolding), TechDocs (Markdown docs as code), Kubernetes plugin, Search, ~150+ plugins (CI/CD / cloud / observability / cost / security).

Why Backstage: most-flexible; OSS; large community; no vendor lock-in.

Trade-offs: significant initial investment (engineer 6-12 months); ongoing maintenance; complex; "you'll spend more on engineering time than you'd pay for managed."

Pick if: 50+ engineers; dedicated platform team; OSS-aligned. Don't pick if: smaller team or want fast time-to-value.

Port — modern managed IDP

Founded 2022. Modern managed alternative to Backstage.

Pricing in 2026: free tier (limited); Pro $1-10K/mo depending on size.

Features: service catalog, self-service actions, scorecards, integrations (90+), Kubernetes / cloud / DevOps tools, Git-as-code (catalog defined in YAML in repos).

Why Port: faster than Backstage to deploy; modern UX; "infrastructure as code" approach; growing fast.

Pick if: want managed without massive engineering investment; modern team. Don't pick if: 100% OSS-aligned.

Cortex — service catalog + scorecards

Founded 2018. Catalog + maturity scorecards.

Pricing: $$ enterprise.

Features: service catalog, scorecards, on-call integration, deployment tracking, eng metrics dashboard.

Why Cortex: best-in-class for "is service production-ready?" scorecards.

Pick if: focus on standards / production-readiness; enterprise / mid-market. Don't pick if: full-platform features needed.

OpsLevel — Cortex alternative

Similar to Cortex.

Pricing: $$ enterprise.

Features: catalog + scorecards + dashboards.

Pick if: alternative to Cortex; enterprise procurement preference.

Roadie — hosted Backstage

Hosted Backstage without infra burden.

Pricing in 2026: $800-3K/mo + per-user.

Features: full Backstage; hosted; Roadie manages upgrades + plugins.

Why Roadie: get Backstage benefits without maintenance.

Pick if: want Backstage but don't want to host. Don't pick if: cost-sensitive (DIY Backstage cheaper if you have time).

Humanitec — platform engineering deep

Platform engineering focus.

Pricing: enterprise.

Features: deployment self-service, environment management, secrets, RBAC, deep K8s.

Pick if: heavy platform engineering; deep K8s.

Mia-Platform — EU

Italian DevOps platform.

Pick if: EU base; Italian / European procurement.

Atlassian Compass

Atlassian-native (Jira / Confluence integration).

Pick if: heavily Atlassian-aligned.

Cloud Dev Environments (CDEs):

Gitpod / GitHub Codespaces / Coder

Pricing: $35-100/user/mo typical.

Features: instant dev environment in browser or local IDE; pre-configured.

Why CDEs: hour-long setup → 30 seconds; consistent environments.

Pick if: dev onboarding is painful; complex setup. Distinct from IDP but complementary.

What IDPs Won't Do

Buying an IDP doesn't:

1. Solve cultural problems. Engineers ignoring standards isn't fixed by tool; needs leadership.
2. Replace platform team. Need 1-3 dedicated FTEs to run IDP at scale.
3. Make engineers self-sufficient. Self-serve actions still require thoughtful design.
4. Scale infinitely. Backstage at 1000+ services starts struggling.
5. Replace docs entirely. TechDocs is great; Notion / Confluence still useful.

The honest framing: IDP is force multiplier on platform team. Without dedicated owner, it becomes another shelfware.

When Are You Ready for IDP?

Decide IDP readiness.

Right time signals:
- 30+ engineers
- 50+ services / repositories
- Multiple teams (3+) with their own services
- Onboarding pain (new engineers can't find anything)
- Outages caused by ownership confusion
- Standardization desired (testing / on-call / monitoring)
- Dedicated platform engineer / team

Wrong time signals:
- <20 engineers (manual works)
- <20 services
- One team / one repo
- Fast-changing architecture (IDP investment goes stale)
- No platform engineering owner

Investment levels:

Light (start small):
- Spreadsheet of services + owners
- Confluence / Notion docs
- Cost: $0; time-bound

Mid:
- Roadie ($800-3K/mo) for hosted Backstage
- 0.5-1 platform engineer
- 3-6 month rollout

Heavy:
- Backstage self-hosted
- 2-5 platform engineers
- 6-12 month rollout
- Custom plugins

For [COMPANY], output:
1. Readiness assessment
2. Right investment level
3. Time-to-value estimate
4. Team requirements
5. ROI threshold (when worth it)

The 30-engineer threshold: below that, manual works. Above 50, you're losing productivity without IDP.

Pragmatic Stack Patterns

Pattern 1: <30 engineers ($0-100/mo)

• Spreadsheet + Notion / Confluence
• Manual ownership tracking
• No IDP yet

Pattern 2: 30-100 engineers ($1-10K/mo)

• Port OR Roadie (hosted Backstage)
• 0.5-1 platform engineer
• 3-6 month rollout

Pattern 3: 100-500 engineers ($5-30K/mo)

• Backstage self-hosted (DIY)
• 2-5 platform engineers
• Custom plugins
• 6-12 month rollout

Pattern 4: Standards-focused ($10-30K/mo)

• Cortex OR OpsLevel for scorecards
• Tracks production-readiness
• Pairs with catalog (Backstage / Port)

Pattern 5: Heavy platform engineering ($enterprise)

• Humanitec for deployment self-service
• Backstage for catalog
• Multi-platform team

Pattern 6: CDE focus ($35-100/user/mo)

• Gitpod OR GitHub Codespaces
• Instant dev environments
• Often complements IDP (not replaces)

Decision Framework: Three Questions

1. Engineering team size?

   • <30 → not ready
   • 30-100 → managed IDP (Port / Roadie)
   • 100-500 → self-host Backstage or managed enterprise (Cortex / OpsLevel)
   • 500+ → DIY Backstage with platform team

2. Primary need?

   • Catalog + ownership → Backstage / Port
   • Standards + scorecards → Cortex / OpsLevel
   • Full platform → Backstage / Humanitec
   • Dev environments → Gitpod / Codespaces (different)

3. OSS vs managed?

   • OSS-priority + engineering capacity → Backstage
   • Speed + simpler ops → Port / Roadie / Cortex
   • Enterprise procurement → OpsLevel / Cortex / Backstage Spotify managed

Verdict

For 30% of B2B SaaS in 2026 reaching IDP threshold: Port for managed modern.

For 25%: Backstage for OSS / engineering-heavy.

For 15%: Roadie for Backstage-without-hosting.

For 10%: Cortex OR OpsLevel for scorecard-led.

For 10%: Humanitec for heavy platform engineering.

For 10%: CDE-only (Gitpod / Codespaces) without IDP yet.

The mistake to avoid: buying IDP at <30 engineers. Tool needs scale; below threshold, doc + spreadsheet works.

The second mistake: Backstage without dedicated platform team. Massive engineering sink; abandoned in 18 months.

The third mistake: conflate IDP with CDE. Different problems; sometimes complementary.

See Also

• Project Management Tools — Linear / Jira / Asana
• Code Quality Platforms — adjacent quality
• Observability Providers — APM
• Error Monitoring Providers — error tracking
• Internal Tool Builders — Retool / Tooljet (different)
• Workflow Automation Providers — Zapier / n8n
• CI/CD Providers — CI/CD tooling
• Container Registry Providers — adjacent
• Application Security Tools — adjacent
• GitHub — Git host
• Workspace Knowledge Base Tools — adjacent docs
• VibeWeek: Internal Admin Tools — adjacent internal
• VibeWeek: Audit Logs — service audit
• VibeWeek: Logging Strategy & Structured Logs — adjacent
• VibeWeek: Metrics & OpenTelemetry Instrumentation — adjacent
• LaunchWeek: Internal Tools Strategy — when to build vs buy
• LaunchWeek: Founder Hiring Playbook — engineering hires
• LaunchWeek: Interview Loop Design — engineering interviews