The Startup That Guarded the Back Door - Then Closed It

There is a peculiar room inside every software company. Customers never see it. Investors rarely ask to see it. The designers who fuss over the public product may pretend it does not exist. Yet this room is where refunds are approved, accounts are unlocked, suspicious users are reviewed, and private records are corrected. It is the admin console - the back door of the modern company - and for years it was commonly assembled from whatever spare lumber engineering could find.

Arisa Amano and Bob Remeika knew the room because they had kept rebuilding it. They worked together at Yammer, then Zenefits, then the digital-securities company Harbor. At each stop, the same awkward request appeared. Customer support and operations needed to reach into production data. Engineers had more visible things to ship. The result was a hurried viewer with broad access, thin controls, and the visual charm of a fuse box.

At Harbor, repetition finally became evidence. If three very different companies required essentially the same hidden machinery, perhaps the machinery was the product. Internal spun out in 2019, with Harbor among its first customers. The founding phrase was “console as a service.” It was unfashionably literal - and unusually good.

A master key disguised as a dashboard

Internal’s first argument was about privacy, not prettier tables. The founders called out the “God View” problem: an employee tool that revealed far more customer information than a particular employee needed. A support representative might need a delivery address and a refund button. A crude console might also expose private notes, payment details, location history, or every other field in the record. Convenience had quietly become permission.

Internal connected to a company’s existing database and read its structure without moving the underlying customer data into a new store. It generated views, then let teams add search, filters, edit actions, custom pages, tasks, and queues. The important part sat underneath: permissions could be applied down to individual fields, sensitive values could be redacted, and actions could be recorded in an audit log. Integrations with systems such as Salesforce, Zendesk, and Google Sheets brought scattered context into one operational view.

“Companies never dedicate enough engineers to their internal consoles, so they’re often half broken and hard to use.”Bob Remeika, co-founder and CTO

This made Internal neither a business-intelligence dashboard nor a general consumer app builder. Dashboards summarize. Internal was meant to let a person safely do something: update a status, review a customer, reset an account, approve a request. Its market sat between direct database access and custom software - close to Retool, and later crowded by Appsmith, DronaHQ, ToolJet, Superblocks, Budibase, and others.

What did the shortcut cost?

Internal sold subscriptions. Its Starter edition cost $0 for one user and included five app users, unlimited custom internal apps, and three data sources. Paid plans charged by user, while larger customers could discuss enterprise and self-hosted deployment. The honest cost comparison was always larger than the invoice: a license bill on one side, and engineering salaries, maintenance, security review, and delayed customer features on the other.

The venture bill is clearer. Craft Ventures led a $5 million seed round in August 2019, joined by Pathfinder - Founders Fund’s early-stage vehicle - and angel investors. An $11 million Series A followed in October 2021, taking reported funding to $16 million.

Customers bought back time. Every.org said it built support tools without engineering overhead. Wonderschool valued a system its product and customer-support teams could shape without surrendering technical power. Dispatch Integration selected Internal for a legacy HR-data project, connected it to PostgreSQL, and delivered secure screens in a few weeks. Another public case study carried the arresting claim that client applications arrived 90 percent faster.

The first thing to fail, in the product’s original diagnosis, was prioritization. Internal tools fell behind customer-facing work, so teams patched together spreadsheets, direct database queries, and admin pages that aged badly. Reviews suggest Internal improved that bargain but did not abolish complexity. Users liked its friendly interface and fast prototyping; some still found setup difficult without knowledge of the underlying database. No-code did not mean no model. Someone still had to understand what the fields meant and what a dangerous action could do.

Forty-three days to find another door

In November 2023, customers learned that Internal would shut down. Competitors quickly published migration pages. Noloco recorded the final date as December 28. From the public notice to the switch-off, customers had roughly six weeks to inventory applications, reproduce interfaces, reconnect actions, retest permissions, and retrain employees.

The sunset changed the product’s central calculation overnight. Services important enough to operate customer support suddenly became migration projects. The irony is sharp. Internal had reduced dependence on scarce in-house engineering; its closure forced customers to summon engineering, vendors, or consultants on a deadline.

That does not make the original decision foolish. A custom internal platform can take months, and it never stops asking to be maintained. A visual builder can be exactly right for a narrow, evolving workflow. The mistake is treating convenience as permanence.

Where the idea still fits

Internal’s approach works best when the data already exists, the workflow is understood, the audience is internal, and speed matters more than a bespoke interface. A support queue, account-review screen, approval tool, or legacy database front end is an excellent candidate. The return comes from avoiding a fresh application stack for a job that mostly consists of finding records and taking controlled actions.

It fits poorly when the application is a permanent core system, requires highly unusual interaction design, serves huge numbers of external users, or must keep running if the vendor disappears tomorrow. Strict portability requirements may favor open-source software or a custom build. Heavy per-seat usage can also reverse the economics that made the first prototype attractive.

Internal’s durable contribution was to take the shabby software behind the company seriously. It saw that an internal console is not merely a shortcut to data. It is a policy made clickable. Every visible field expresses a decision about trust; every button expresses a decision about power. The company closed, but that observation did not. The back door still deserves a designer, a lock, and - as Internal’s customers learned in 2023 - a fire exit.