Vendor cost framework

Cortex IDP cost in 2026: scorecard-first buyer framework

Cortex makes engineering-standards measurement the primary object of the IDP. The cost-justification is straightforward if scorecards are a strategic priority and harder if they are not.

AWS Marketplace list

$39,000 / yr

Published SKU: 50 Cortex IDP users, SaaS hosted, 12-month contract. Checked 26 Sep 2026.

Implied per user

$780 / user / yr

$39,000 divided by the 50 users in that SKU. Cortex publishes no general rate card on cortex.io.

Longer contracts

Up to 46% off

Marketplace lists up to 40 percent off on 24 months, up to 46 percent on 36 months. Overage $1.00 per user-hour.

What scorecard-first actually means

Most Internal Developer Platforms (Backstage, Port, OpsLevel, Compass) treat the service catalogue as the primary object: the IDP knows about your services, owners, dependencies, deployments, and uses that knowledge as the substrate for everything else. Scorecards (automated rubrics that grade services on reliability, security, ownership, documentation, on-call discipline) are an important feature, but they are a feature.

Cortex inverts that ordering. The catalogue is still there and is still the substrate, but the primary product surface is the scorecard. The implicit thesis is that engineering leaders most want their IDP to drive engineering quality (consistency of standards, observable improvement, accountability across teams) and the right primary tool for that is automated rubrics with dashboards and team-level reporting.

If that thesis matches your engineering leader's mental model, Cortex is the cleanest expression of it on the market. If your IDP goal is self-service infrastructure first or developer portal first and scorecards are a distant priority, the case for Cortex specifically (over Port or Backstage) is weaker on a feature basis even if the cost is comparable.

Pricing model

Cortex's own pricing page does not publish a rate card. It describes a per-seat model that scales with organisation size and routes buyers to a quote. So the honest starting position is that Cortex is a quote-based vendor, and anyone quoting you a confident Cortex annual figure for your org size is guessing.

There is, however, one genuinely public Cortex price, and it is worth knowing before you take a first call. Cortex lists on AWS Marketplace, and that listing publishes dimensions rather than hiding behind a private offer. Checked on 26 September 2026 it shows:

$39,000 per 12 months for the dimension listed as 50 Cortex IDP users, SaaS hosted. Dividing by the 50 users in that SKU gives $780 per user a year, or $65 per user a month.
Term discounts of up to 40 percent on a 24-month contract and up to 46 percent on a 36-month contract.
An overage fee of $1.00 per user-hour for consumption beyond committed capacity, which is the mechanism that bites if your seat count drifts upward mid-term.
A second dimension the listing itself labels a depreciated SKU at $100,000 per 12 months. That is a legacy line, not a current tier, and it should not be read as Cortex's enterprise price.

What that does and does not tell you: it is a real list price on a public surface, so it is a far better anchor than a band someone invented. It is not a rate card. Cortex says pricing scales with org size, and a 100-seat deployment is not simply two of these. Doubling the SKU gives $78,000, but that arithmetic assumes linear per-seat pricing that Cortex does not publish and that volume discounting normally contradicts, so treat $78,000 as the top of a sanity-check range rather than an expected quote. If you need a number for a 100-engineer budget line, ask Cortex and use the $780 per user list rate as the ceiling you are negotiating down from.

What you avoid building by buying Cortex

The build-equivalent of Cortex is roughly:

A scorecard engine with a rule DSL, scheduled evaluation, history, drill-down by team and service. About four to six engineer-months of build for a credible MVP.
Integrations with common data sources (Git, CI, observability, incident management, secrets, identity). Each integration is one to three engineer-weeks; the common-case set is six to ten integrations.
A team-level reporting surface with dashboards, trends, and Slack or email digests. About two to four engineer-months for a credible MVP.

Total avoided build work is roughly 12 to 24 engineer-months in year one, or $234k to $468k of avoided platform-engineer cost at the senior loaded rate (see /salary). Against that, the $39,000 published 50-user list price (or $78,000 if you scale it straight to 100 seats) fits comfortably inside the avoided-build window. That is the core buy-side argument for Cortex and it survives the uncertainty about your specific quote: even at several times the published rate, the licence is cheaper than building a scorecard engine.

What you still staff for after buying Cortex

Three lines stay on the platform team's plate. Treat them as the real annual cost of running scorecards, with Cortex providing the substrate.

Data-source integrations. The stock connectors cover common cases (the major Git providers, PagerDuty, the major CI providers, the major observability vendors). Your unusual sources (the internal billing API, the homegrown deployment system, the proprietary alert manager) need custom integrations, typically one to three engineer-weeks per source.
Scorecard rule authoring and tuning. A scorecard with five rules that mostly evaluate true is useless; a scorecard with twenty rules where everyone scores poorly is demoralising and gets ignored. The right rubric (typically eight to fifteen rules per scorecard, calibrated so that the median team scores around 70 percent and improvement is visible quarter to quarter) takes 4 to 8 engineer-weeks to land for year-one and another 2 to 4 a year for ongoing tuning.
Rubric ownership and defence. When a team scores poorly the platform team has to either help them improve or accept that the rubric is wrong. Caving to "make the rubric easier" is the failure mode that kills scorecards. Owning the conversation requires a senior platform engineer or manager with credibility and time, about 0.25 FTE long-term.

Three-tier deployment cost

The licence half of this we can only anchor to the published 50-user SKU. The rest is our own model of the rubric and integration work that lands around it, at the $234,000 loaded senior rate (see /salary).

Lean. Cortex at roughly the published 50-user rate ($39,000 at list), plus stock integrations and one scorecard on the default rubric. Our model adds a quarter to a half of a platform engineer, so budget about $100,000 to $160,000 all in. Suits a 50-engineer organisation where the goal is knowing who owns what and whether services have runbooks.
Standard. A 100 to 250-seat deployment, which is past the published SKU and therefore quoted. Plus 4 to 6 custom integrations, three to five scorecards, and quarterly rubric reviews. Our model sizes the surrounding work at one to one and a half platform engineers, roughly $234,000 to $350,000 of loaded cost, on top of the licence. Most Cortex customers sit here.
Enterprise. Licence is quoted and we do not estimate it. The surrounding programme (FinOps and incident-management integration, team-level OKRs tied to scorecards, dedicated rubric ownership at platform-engineering manager level) is what our model can size: two to three platform engineers plus management time, roughly $470,000 to $700,000 a year. Suits a 500+ engineer organisation with a strong engineering-standards culture.

Crossover with Port and Backstage

The five options in this category do not disclose equally, so "roughly the same licence band" is not a claim the evidence supports. Three publish something: Cortex at $780 a user a year implied by its Marketplace SKU, Port at $480 a user a year on Standard, Roadie at $288 a developer a year on Teams. Atlassian Compass publishes the cheapest rate card in the category by a wide margin at $7.67 per user a month on Standard, or $92 a year (see /compass-cost). OpsLevel publishes nothing at all. On published rates the spread across the category is roughly eightfold per seat, not a narrow band, though Compass is a lighter product than the rest and the comparison is not strictly like for like. The feature-fit differentiator is still usually the deciding one, but do not assume the licence cost washes out.

The clearest cost-vs-substrate trade-off is against self-hosted Backstage. At year one, Cortex beats self-hosted Backstage on cost because you avoid the platform-engineer headcount needed to install and customise Backstage and the scorecard plugin work to get to feature parity. At year three or later, the maths can flip if the self-hosted Backstage deployment is mature and the platform team is large enough to absorb operations. Crossover sits around 300 product engineers for most orgs.

The clearest cost-vs-feature trade-off is against Port. On published list rates Cortex is the dearer of the two per seat ($780 a user a year implied, against Port's $480 on Standard), so the question is whether scorecard depth justifies the premium, or whether flexible self-service actions are the higher IDP priority. If scorecards, Cortex. If self-service actions and a flexible entity model, Port. Many large organisations end up with both, which doubles the licence cost but is sometimes the right answer (see /build-vs-buy for the framework).

When Cortex is the right pick

Cortex is the right pick when all of the following are true:

Engineering leadership treats standards measurement (scorecards, rubrics, team-level reporting) as a strategic priority.
You have at least 50 services and at least three meaningful standards to score (ownership, reliability, security at minimum).
Your platform team has the capacity to integrate Cortex with at least the major data sources and to author and defend the rubric long-term.
You can absorb the per-year licence into your platform budget without it crowding out tooling spend elsewhere.

Outside that window, the answer is one of the other commercial IDPs (/port-cost, /opslevel-cost, /compass-cost), one of the Backstage routes (/backstage-cost, /backstage-hosted-cost), or "no IDP yet, lightweight checklist culture for now" if your org is under 30 engineers.

Prices quoted from the Cortex AWS Marketplace listing ($39,000 per 12 months for 50 SaaS-hosted users, term discounts, $1.00 per user-hour overage) and the Cortex pricing page (per-seat, scales with org size, quote only), both checked live 26 September 2026. Marketplace list price is a public fact; what Cortex sells at is not something we know. Per-user and 100-seat figures are our arithmetic over that published SKU, clearly labelled where they appear. Comparison rates are from the Port, Roadie and Atlassian Compass pricing pages, checked the same day. Engineer-time figures are our own model at the loaded rate on /salary.

Frequently asked questions

How much does Cortex cost?

Cortex's own pricing page publishes no rate card: it states a per-seat model that scales with org size and directs buyers to a quote. But Cortex does list a real SKU on AWS Marketplace, and that listing is public. As checked on 26 September 2026 it publishes $39,000 for 50 Cortex IDP users, SaaS hosted, on a 12-month contract, which works out at $780 per user a year. The listing also publishes longer-term discounts (up to 40 percent off on a 24-month contract, up to 46 percent on 36 months) and an overage fee of $1.00 per user-hour beyond committed capacity. That $39,000 figure is the only public Cortex price point we can point at, and it is a Marketplace list price rather than what Cortex necessarily sells at.

What is the scorecard-first IDP value proposition?

Most commercial IDPs treat the service catalogue as the primary object and add scorecards as a feature. Cortex inverts that: scorecards are the primary object, the catalogue is the substrate. The implicit thesis is that engineering quality (reliability, security, ownership, documentation discipline) is the thing leaders most want their IDP to drive, so the tool should optimise for measuring and improving it. The cost-justification therefore lands on how much you value automated scorecard reporting on standards like service ownership, on-call coverage, runbook freshness, and security posture.

What do Cortex scorecards need from the platform team?

Three things. First, data-source integrations: scorecard rules need to read from the systems where the underlying signals live (Git, PagerDuty, the CI provider, the secrets manager, the observability stack). Stock integrations cover the common cases; unusual systems need a custom integration. Second, scorecard rule authoring: defining the rules that actually matter (and tuning them so they do not generate scorecard fatigue) is platform-team work. Third, ownership of the rubric: when a team scores poorly, someone has to make the case for why the rubric is right, not soften the rubric.

How does Cortex compare to Port and Backstage on cost?

Only two of the three publish anything, so a like-for-like comparison is partial. Port publishes $40 per seat a month on Standard, or $480 a user a year. Cortex's AWS Marketplace SKU implies $780 a user a year at the 50-user list price. Roadie publishes hosted Backstage at $24 per developer a month, or $288 a year. On published list rates Cortex is the dearest per seat of the three and Roadie the cheapest, though all three discount on negotiation and Cortex's figure is a 50-user SKU rather than a general rate card. Self-hosted Backstage swaps subscription cost for headcount. Feature fit usually decides this rather than list price: Cortex is differentiated by depth on scorecards, Port by the flexible entity model and self-service actions, Backstage by openness and the plugin ecosystem.

What is Cortex not the right choice for?

Cortex is not the right choice when scorecards are not a strategic priority for engineering leadership. If the IDP goal is self-service infrastructure or developer portal first and engineering-standards measurement is a distant second, Port or hosted Backstage are usually a better fit. Cortex also is not the right choice if your organisation is small enough (under 30 engineers) that lightweight checklist culture works as well as automated scorecards; scorecard infrastructure has fixed setup cost that needs at least 50 services to pay back.

What is the typical year-1 deployment cost of Cortex?

Here is the arithmetic rather than a band. Licence: the only public anchor is the AWS Marketplace SKU at $39,000 for 50 users, so a 100-seat deployment is somewhere between that and $78,000 at list, and is quoted rather than published. Platform-team time: 6 to 10 engineer-weeks for initial integration setup, 4 to 8 for the first round of scorecard rule authoring, 4 to 6 for adoption work (rollout, training, soft launch). That is 14 to 24 engineer-weeks, which at the $234,000 loaded senior rate on our salary page is about $63,000 to $108,000. Adding the two gives a year-one total of roughly $100,000 to $190,000, of which the licence is a third to a half. After year one the steady-state operations cost typically drops to 0.5 to 1 platform engineer of ongoing work.

Port IDP cost

Flexible entity model alternative.

OpsLevel cost

Catalogue-plus-maturity alternative.

Self-hosted Backstage cost

Open-source alternative.

Hosted Backstage cost

Hosted alternative.

Build vs buy

The framework all this feeds into.

Run your numbers

Plug your dev-count into the calculator.