Ramp and Stripe each built an internal vibe-coding tool with four engineers or fewer
Ramp built Glass, an internal vibe-coding platform where non-engineers make apps on connected company data, the Pragmatic Engineer newsletter reports. Stripe's comparable tool, Harbor, took one engineer about six weeks. Connector access is the open question.
The Engineer · Build desk
A single engineer built Stripe's Harbor in around 6 weeks. Ramp had 2 engineers put Glass together in a matter of weeks; the newsletter also describes a four-person development team behind Glass.
| Figure | Value | Note | Claim |
|---|---|---|---|
| Time for a single engineer to build Stripe's Harbor | 6 weeks, around | 5 | |
| Ramp engineers who put Glass together | 2 engineers | in a matter of weeks | 6 |
| Glass dev team at Ramp | 4 people | 1 |
What happened
- One popular Glass app is a custom CRM built by someone in marketing, with more than 50 fields and filters on screen.
- Ramp and Stripe both made the generated apps follow internal design guidelines, which the author says third-party tools handle poorly out of the box.
- Ramp has since launched Glass as an external product, billed as an AI workspace where knowledge compounds.
Why it matters
- Who's exposed exposure Any internal system wired in as a connector is reachable from apps that non-engineers write and share with colleagues.
- What it costs cost Harbor v1 cost Stripe roughly six engineer-weeks, a small build bill; the recurring cost of reviewing connectors is not reported.
- What doesn't add up contradiction The newsletter puts Ramp's Glass at a four-person team in one place and two engineers building it in weeks in another, so the cost figure depends on which count you use.
In Glass, internal systems are wired in as "connectors", and those become the data sources for the apps people build [3]. The connector is therefore where data access gets decided. There are two ways to build it. A connector can act as the end user, so a marketer sees only what a marketer may see. Or it can act as a service account, so every app on it sees whatever the account sees. The newsletter does not say which design Glass uses. Apps can be shared for others to use and modify [2], so under the service-account design a shared app is a path to the data for anyone who opens it.
Platform teams moving from building every tool to governing connectors is a reasonable inference from this setup, and we hold it as an inference. The newsletter's account of roles stops at engineers who build the tool and non-engineers who build apps on it [2]. If a team of four [1] or a single engineer [5] can stand up the builder, the recurring work left on the platform side is the connector list and the rules on each connector. Apps then multiply without a ticket, while each connector stays a standing decision.
Stripe calls Harbor an "AI-assisted prototyping tool", a label the author says is more elegant than his own, vibe-coding tool [5]. His case for building in-house is a list he says he gathered, not a set of statements from Ramp or Stripe [10]. The data-safety point in that list holds because the apps are internal.
The investment framing has a condition. The author ties his expectation to tech companies that already run their own coding agent harness in their cloud [12]. Without that harness, six engineer-weeks is not a transferable figure. For those companies he wrote that "it's not that complicated, gets much higher adoption than vendor products, and helps build out more AI infra expertise in-house" [14].
The adoption evidence is Ramp's. Onboarding to Glass and building a simple tool is part of new-hire onboarding there, and pretty much every non-engineer uses it [8]. That is the experience of one company.
What to watch
- Whether Ramp documents how Glass connectors authenticate, as the end user or as a service account, now that Glass is sold externally.
- Whether other mid-sized companies the author says are building similar tools publish their connector and permission designs.
- Whether Ramp or Stripe release usage counts for Glass or Harbor to back the adoption claims.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence35
- Adoption40
- Hype gap+25
- Incentives40
- Confidence55
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
The author visited Ramp's headquarters in New York and talked with the four-person development team behind 'Glass', Ramp's 'internal Lovable'.
- [2]
Ramp's Glass is a vibe-coding platform for non-engineers to create complex applications which run internally, and users can share the applications they make with others to use and modify.
- [3]
Users can connect various internal systems as 'connectors' that can then be used as data sources for their apps.
- [4]
Among popular Glass apps was a custom CRM built by someone in marketing, which had 50+ fields and filters on-screen.
- [5]
Stripe's tool, called 'Harbor' and described by Stripe as an 'AI-assisted prototyping tool', was built by a single engineer in around six weeks; the author calls it a vibe-coding tool and says so is his own wording.
- [6]
The author's list of reasons says Ramp had two engineers put Glass together in a matter of weeks, and Stripe had a single engineer build Harbor v1.
- [7]
Ramp has launched Glass as an external product, describing it as 'An AI workspace where knowledge compounds.'
- [8]
At Ramp, onboarding to Glass and creating a simple tool is now part of new-hire onboarding, and pretty much every non-engineer uses Glass to build their own tools.
- [9]
Stripe's Harbor v1 took roughly six engineer-weeks of effort.
- [10]
Among the reasons the author says he gathered for building in-house: because the tool produces internal-facing tools it is a lot safer to expose internal data sources to it, and doing the same with third-party tools like Lovable or Replit would be risky.
- [11]
Both Ramp and Stripe paid special attention to ensuring the generated apps follow internal design guidelines, which the author says third-party apps are not great at out of the box.
- [12]
Among tech companies that have built their own coding agent harness that runs in their cloud, the author expects many to build internal vibe-coding tools for non-engineers like Stripe and Ramp have done.
- [13]
The author has learned that many other mid-sized-and-above companies are building similar 'internal Lovables' that don't usually see the light of day in the outside world.
- [14]
Gergely wrote: "it's not that complicated, gets much higher adoption than vendor products, and helps build out more AI infra expertise in-house."
Get this ledger inside ChatGPT. Ask about any of our stories and the answer brings its figure, each claim’s verdict and the sources. Free, no account.
Sources
1 independent publisher whose own reporting we read for this story.
- blog.pragmaticengineer.comThe Pulse: new trend of building “internal vibe-coding apps” at tech companies
1 article · October 15, 2026
- newsletter.pragmaticengineer.comThe Pulse: new trend of building “internal vibe-coding apps” at tech companies
1 article · October 15, 2026
Topics and entities
Follow any of these and your For You feed starts watching them — no settings page required.
Topics
- Internal toolsFollow
- Vibe CodingFollow
- Build vs BuyFollow