The quiet plumbing under your favorite company's admin panel. The reason a Stripe engineer goes home for dinner instead of writing yet another CRUD app.
At any given moment, somewhere inside Amazon, DoorDash, the NBA, Brex, Lyft, and roughly a third of the Fortune 500, a customer-service agent is staring at a screen that does exactly one thing. Refund a charge. Re-route a driver. Reset a vendor's password. Pull a contract. The screen is ugly. The screen is loved. The screen was built in Retool.
Retool, the San Francisco company founded in 2017 by a 25-year-old named David Hsu, sells what most engineers privately consider the world's most boring software: internal tools. CRUD apps. Admin panels. Workflows that fire at 3am. The kind of thing that never shows up on a careers page and never wins a design award. By 2022, that boring software had earned the company a $3.2 billion valuation and a customer list that reads like a guest book at Davos.
Talk to any engineer at a fast-growing startup and they will, eventually, get to the part where they sigh. The sigh arrives at the mention of the admin panel. Of the ops dashboard. Of the workflow that someone in Finance asked for six weeks ago and which, when delivered, will be deprecated almost immediately by a new ask.
Hsu noticed this in the most painful way possible: by being the engineer who had to build it. Before Retool, he ran a UK-based Venmo competitor called Cashew through Y Combinator's Winter 2017 class. The consumer payments business was, in the polite British formulation, not working. The company was losing money on every transaction. The runway was sixty days. Most founders would have quietly returned the keys.
Hsu and his small team looked instead at what they had actually built. The user-facing app was bleeding cash. The internal tools, however - the dashboards, the refund flows, the customer-service consoles - were genuinely useful. Not as software. As scaffolding. They had built the same patterns half a dozen times.
Low-code platforms were not new in 2017. They were, by then, a graveyard of well-intentioned products that promised non-technical users could build real software and then very politely failed to do so. Retool's bet was sharper, and frankly stranger.
The bet: developers actually like code. They just hate writing the same JavaScript-and-SQL CRUD scaffolding for the four-thousandth time. Give them a drag-and-drop UI for the parts that bore them, give them an editor and full programming languages for the parts they care about, and they will choose you. Not under duress. By preference.
This sounds obvious now. In 2017, it sounded like a category error. Most low-code tools were chasing the citizen-developer dream. Retool turned around and aimed at the people the citizen-developer movement had implicitly written off: engineers with strong opinions and a deadline.
By the time Retool went into general availability, the company was reportedly already at roughly $2M in annual recurring revenue. Word of mouth from one engineer to another, on Slack and at lunch, did the marketing.
Seven years, five products, one persistent thesis. Below: the milestones, mostly without the press-release polish.
Retool today is a suite, not a single tool. The original drag-and-drop app builder is still the front door, but the company has spent the last few years methodically filling in the rooms behind it.
The original. Drop a table, wire it to Postgres, write a SQL query, ship to ops by Friday.
Cron jobs, webhooks, and human-in-the-loop steps. The engine for everything that should not need a Zoom call.
Native iOS and Android internal apps. For drivers, warehouse staff, field technicians, and other people who refuse to use laptops.
LLM connectors, vector databases, and an agent framework. Anthropic and OpenAI under the hood. Guardrails optional but recommended.
A managed Postgres for the times you have an idea before you have a backend.
For banks, hospitals, and anyone whose lawyers have opinions about where the data sleeps at night.
Retool's funding pattern is the unusually patient kind: the company waited two and a half years between seed and Series A, then stacked rounds quickly once the wedge was obvious. The chart below is not a forecast. It is a paper trail.
Officially, Retool's mission is to help developers and companies build the internal software they need to run their business, faster and with less drudgery. Unofficially, you can read it as a polemic. The polemic is that the world's engineering talent is, in aggregate, badly allocated. Too many of the smartest people in any given building are rebuilding the same five admin patterns, and they know it.
That insight is now being extended into the AI era, which is, depending on the day, either an opportunity or an existential threat to a category called low-code. Retool's answer has been to absorb it. The same drag-and-drop primitives now wire LLMs, vector databases, and agents into the same boring, useful internal tools.
Customer support consoles. Inventory dashboards. KYC review queues. Fraud-investigation workbenches. Refund pipelines. Trip dispatch panels at Lyft. Content-moderation tools. Hiring tools. The unglamorous, load-bearing software that keeps companies upright.
Most low-code platforms ask developers to give up the keyboard. Retool asks them to keep it, but use it less often. The difference is the entire pitch.
Amazon. DoorDash. Stripe. NBA. Pinterest. Brex. Coinbase. Mercury. Lyft. Volvo. Philips. Plus, somewhere out there, a five-person startup running its entire back office on a free-tier Retool instance.
Return, for a moment, to that customer-service agent at Amazon. The screen is still ugly. The screen is still loved. But the screen, increasingly, is not just refunding a charge. It is summarizing the customer's last six tickets with an LLM, suggesting a remedy, drafting a response, and asking the human to approve. The drudgery has moved up a level. The agent has been promoted.
That is the long bet Retool has been making since 2017, before it had words for it. Internal software is the shape companies actually take. The faster you can mold that shape, the more company you can be. AI does not replace the admin panel. It makes the admin panel the place where decisions get made.
Retool started as a way to skip the boring parts. It is now, quietly, a way to industrialize them. Somewhere, a developer is dragging a table onto a canvas, wiring it to a database, and going home. Somewhere else, a vice president is staring at a dashboard he did not commission, built by an engineer he has never met, that tells him exactly what is wrong with his quarter. Both screens, in 2026, are Retool.