Scuba Raises $34M Across Pre-Seed and Seed for AI Context

Accel has already put Scuba to work inside its own team. The investor says it uses the shared-context product every day, giving a young company's financing a useful operating detail: people allocating capital are also encountering the software in the course of their work.

Scuba, a New York startup founded by CEO Noah Desai Weiss and CTO Dolapo Falola, emerged from stealth on October 13, 2026, with $34M across pre-seed and seed rounds. Accel led the pre-seed and co-led the seed with IVP, alongside participation from BoxGroup, Conviction, Homebrew and Scribble Ventures. The company is opening limited beta access to teams in waves.

The product brings saved information into shared Collections that employees and AI agents can use. Scuba's opportunity sits in the movement of judgment through a company: a useful customer conversation or decision should still matter when a different colleague, or a different agent, picks up the work. Building around that movement gives the financing a more specific subject than another application promising faster answers.

What Scuba's $34M financing covers

Scuba's launch announcement describes $34M over 2 rounds, with distinct lead roles for the pre-seed and seed. Individual round amounts, valuation and closing dates were not disclosed. The announced total therefore belongs to the combined financing history, rather than a single $34M seed round.

The founders bring a long working relationship to the company. Noah Desai Weiss was Slack's CPO, and Dolapo Falola previously led engineering at The Browser Company. Scuba's own account places their collaboration back in 2011, with careers through Google, Foursquare and Slack. Their new company is developing software for a coordination problem they have had years to encounter inside product organizations.

Accel's investment account, published on October 13, connects its backing to that history and to its own use of the product. The firm sees value in bringing organizational decisions, conversations and processes into AI work. Its participation supplies an informed early audience, while paid-customer count, revenue and retention remain undisclosed. Those commercial measures will matter as beta access broadens beyond the early users already experimenting with Scuba.

How Captures become shared organizational context

Scuba's product description starts with information people choose to save. Articles, screenshots, documents, videos and threads become Captures; Collections group shared material around a launch, client or research subject. Scuba describes automatic organization, summaries and responses with links back to the underlying material.

That arrangement gives the employee an editorial job. Saving a customer conversation preserves something the team may need later, but sharing it also determines who can use it. An agent drafting a proposal has a better chance of carrying the team's reasoning forward when the relevant evidence is available. Selecting and maintaining that evidence becomes part of the work surrounding the draft.

The company describes MCP connections that let Claude, ChatGPT and Cursor read and create Captures, alongside a Slack integration for questions and thread capture. Those connections carry Scuba's context into other tools rather than requiring every task to happen inside its own interface. For a buyer, the practical question is whether the material reaches the colleague or agent at the point where it changes a decision.

Scuba's founders report hundreds of early users over 6 months of private beta. That is an early adoption signal, with no disclosed breakdown of paying accounts or sustained usage. The company is now widening a limited beta, leaving the product's broader commercial shape to develop through the teams joining it.

Sharing judgment requires permission choices

Scuba's security and privacy page gives the shared-memory proposition a boundary. Captures are private by default, and users decide what to share. The company says it saves chosen material, supports optional auto-capture for selected websites and has no passive screen, microphone or webcam recording.

Those details matter because organizational context can contain more than a useful fact. A meeting record can also carry a customer's concern, an employee's unfinished thinking or information meant for a smaller group. Making that record convenient for an agent creates a reason to understand the permissions around it. A context product has to preserve the intended audience as it makes information easier to reuse.

Scuba says data is encrypted in transit and at rest and that neither it nor its AI vendors train models on customer data. Its SOC 2 audit is underway, according to the current security page. Procurement teams evaluating the beta have a concrete distinction to work with: stated controls and an audit in progress, with certification still to be completed.

The wider operating issue is ownership of shared context. Somebody must decide when material has become obsolete, which Collection should hold it and who should have access. Automatic organization can make those decisions easier to carry out, while the organization still supplies the judgment about what belongs in the record.

A small team building for the next handoff

Scuba's company page describes a ten-person team working together in New York's West Village, 5 days a week. Much of the team has experience at Google, Slack, Meta and growing startups. That physical arrangement gives the company a close setting in which to develop a product intended to carry context beyond an immediate conversation.

The company is offering early access and links to its careers page. No specific hiring target or detailed allocation of the financing has been verified. The visible product agenda is the limited-beta rollout and the work of bringing more teams' material into Captures and Collections.

For operators, Scuba makes the quality of organizational memory part of an AI purchasing conversation. Useful answers depend on what the system can find, how recently the team maintained it and whether the person receiving an answer can trace its reasoning. The buyer has to evaluate the software alongside the habits it asks a team to adopt.

As more teams enter the beta, Noah Desai Weiss and Dolapo Falola will encounter contexts their own ten-person organization never had to manage. A client's unusual request, an old decision that suddenly matters again and a colleague picking up someone else's work will give those teams reasons to decide what goes into a shared Collection, and what should travel with it.