Traefik Labs

Here is a fact that sounds like a joke but is a real pitch a real company used to raise real money: Traefik Labs wants to make networking boring.

Networking, to be clear, is the plumbing that gets a user's request from their browser to the correct piece of your software and back again. When it works, nobody notices. When it breaks, everybody notices, usually at 3 a.m., and someone gets paged. So "boring" here is not an insult. Boring is the highest compliment infrastructure can receive. Boring means the pipes held.

The origin story is the kind engineers find deeply relatable. Around 2015, a French engineer named Emile Vauge was deploying thousands of microservices - lots of small programs that all need to find and talk to each other - and the existing tools for wiring them together required writing enormous amounts of configuration by hand. Every new service meant more config. More config meant more places to be wrong. So Vauge did the thing engineers do when a tool annoys them enough: he wrote his own.

The result was Traefik Proxy, a reverse proxy and load balancer with one genuinely clever trick. Instead of making you tell it where all your services live, it watches your infrastructure - your Docker, your Kubernetes - and figures the routing out itself. You launch a container; Traefik notices and starts sending traffic to it. You kill the container; Traefik stops. The configuration you did not write is configuration you cannot get wrong.

This is a small idea with a large blast radius. Traefik Proxy has now been downloaded more than three billion times. It has north of 60,000 stars on GitHub and a community of over 750 contributors, and it runs on more than 100,000 instances doing actual work for companies including Condé Nast, Expedia, Mailchimp, and Mozilla. It is one of Docker Hub's top-15 projects and has been ranked the number-one open-source API gateway six years running. For a piece of software written in Go and given away free under an MIT license, that is an enormous footprint.

The "given away free" part is the business question, and Traefik's answer is the open-core model. The proxy is free, which is how you accumulate three billion downloads and become a default. Then, on top of the thing everyone already trusts and runs, you sell the things companies will pay for: Traefik Enterprise and Traefik Hub, which add API management, high availability, security, autoscaling, and - crucially for the person signing the check - support. The free product is not charity; it is the world's largest distribution channel.

In 2016 Vauge founded a company, Containous, to steward the project, and later renamed it Traefik Labs so the company and its most famous product would answer to the same name. It raised a $10 million Series A in January 2020, led by Balderton Capital with Elaia, 360 Capital, and Kima Ventures along for the ride - roughly $11 million total across its life, which by the standards of modern infrastructure fundraising is a notably restrained number. Traefik built a category-leading open-source project and a commercial business on top of it without a nine-figure war chest, which is either a constraint or a moat depending on how you feel about competitors who cannot buy their way to three billion downloads.

Then, in February 2024, the founder did something founders rarely do voluntarily: he replaced himself. Vauge handed the CEO title to Sudeep Goswami, who had joined the previous year as chief revenue officer, and moved into the CTO seat to get back to the engineering he started with. "Announcing a new era," he wrote at the time. It is a quietly unusual move. The mythology of startups says the founder-CEO clings to the chair. Vauge decided the right person for the growth stage was someone else, and that he would be more useful writing the roadmap than managing the org chart.

Which brings us to the part of the story where the plumbing meets the moment. The current frontier for a traffic-routing company is, inevitably, AI traffic - and Traefik has leaned in with what it calls the Triple Gate: one gateway, three kinds of traffic. There is the API Gateway you would expect. There is an AI Gateway that turns any large-language-model endpoint into a managed, rate-limited, cost-controlled API. And there is an MCP Gateway, which governs what AI agents are actually allowed to touch when they go rummaging through your systems via the Model Context Protocol. If you have ever worried about an autonomous agent doing something expensive or unwise, this is a gate you put in front of it.

The through-line from 2015 to now is remarkably consistent. Traefik's whole reason for existing is to sit in the middle - between users and services, between apps and APIs, and now between agents and everything else - and to make that middle a place where governance happens and mistakes do not. The traffic keeps changing. The job of quietly, reliably routing it does not. Boring, it turns out, scales.