The case for Meerkat

Every professional application needs the same foundation before its first real screen: a polished sign-in, two-factor, single sign-on for your customers, roles, an audit log, certificates, monitoring. Your buyers check all of it. None of it makes your product different. And all of it takes time to build, then more time to maintain.

In short:

• One product instead of eight, and one pod instead of about 38.
• 5.5 to 12 days to set up, instead of 83 to 165.
• $27,000 to $88,000 a year: the licences of the equivalent assembled from paid services.
• Free and complete in its Community edition.

What buyers expect in 2026

This is the list as it appears in security questionnaires and tenders.

What Meerkat covers, line by line, is on what it does.

What you assemble without it

For each need, the product teams usually pick: a free one they host themselves, or a paid service.

Two needs have no product at all: a user menu and portal inside your own applications, and one audit log across every tool. Without Meerkat, you write those yourself.

What it costs to run

Every product you add brings its own servers, database, backups, upgrades and security alerts. Here is the same front door built both ways, with the number of pods each product's documentation recommends for production.

Assembled, that is about 38 pods and five storage systems to install, secure, upgrade and back up.

With Meerkat: one pod, 27 MB at rest and 48 under full load. Three pods and a PostgreSQL when you want high availability.

And it is fast

One product instead of eight is only good news if that product is not the slow one. So we measure it against Kong, APISIX and Traefik: each limited to the same single CPU, in front of the same service, under the same load, in the same run. The table compares products, not machines. The results are recomputed on every change to the code and shown live: the measurements.

What it costs in licences

Free products cost nothing in licences: you pay for them in servers (above) and in time (below). Paid services show a price. To compare them, take a software vendor selling to businesses, the case where this foundation weighs the most:

• 5,000 active users a month
• 20 customer organisations, 10 of them with their own SSO
• 15 services, in production and staging
• 50 million requests a month
• a team of 10 developers

With 20 organisations, this case needs Meerkat Enterprise.

Public list prices in US dollars, before tax, without servers or integration work. Products with no public price for this case are left out rather than guessed (Kong Konnect beyond 10 million requests, Tyk, Kong Enterprise, Pomerium Enterprise, Frontegg beyond five connections). For scale: Traefik Hub sells for $30,000 to $50,000 a year, and per-user offers such as Cloudflare Access or Pomerium Zero, at $7 a user, would reach $35,000 a month for 5,000 users.

Each customer that brings its own SSO

Identity services bill single sign-on per customer. So the foundation costs more with every large customer you sign - exactly when you are winning.

Connections included before the meter starts: five at Stytch and Descope, three on Auth0 Essentials, one at Clerk, none at WorkOS. After that, every customer arriving with its Entra ID or Okta is billed, whatever it uses. With Meerkat, a customer's SSO is a setting, not a line on the invoice.

What it costs in time

The heaviest cost is not the licence, it is the engineering. Here is our estimate, in person-days, to reach the same result, from the low estimate to the high one. Adapting your own services to read the user's identity is left out: every option needs it.

SaaS saves installing servers, but not connecting the products to each other, nor building what none of them covers.

Where these figures come from

Measured. Memory and speed come from the project's own benchmark (tools/bench in the repository), last run on 18 October 2026 on GitHub's x64 and arm64 machines, at rest and under load. What is built is read from the code, not from a plan: the public inventory counts today 187 features delivered, 53 in part and 15 to come.

Gathered. Prices are the public list prices, in US dollars before tax, read on the official pages on 16 September 2026, without any discount. A buyer who negotiates will pay less, which is why the comparison uses list prices: they are the only ones anyone can check. Pod counts and memory are the ones each product's documentation gives for production.

Estimated. The person-days are Softwarity's estimate, given as a range because that is what it is, converted at 650 EUR a day before tax for an experienced DevOps or security engineer.

Prices change. If a line here is out of date, tell us and we will fix it.

What you get instead

One product instead of eight. One pod instead of thirty-eight. A foundation that is already there, instead of one to build.

For the team that lives with it:

• One thing to run. One image, one console, one audit log, one backup. Not eight products to keep up to date, each with its own security alerts.
• Ready for the security questionnaire on day one. Two-factor, passkeys, single sign-on over OpenID Connect, audit, TLS, vault: all in the free edition. Answering a tender stops being a project.
• Your developers back on your product. Five to twelve days of setup instead of 83 to 165, and above all two to five days of upkeep a year instead of twenty to forty, every year.
• No lock-in. The Community edition is complete and will stay so, and the code becomes Apache 2.0 after two years. You are not betting your front door on us.

What is built and what is not is in one table read from the code. What each edition includes is on editions. And to see it before deciding, the showcase has the screens.