On a single June day, Niteshift did what most startups spend months choreographing: announced both its seed funding and the general availability of its product. The New York-based startup pulled in $7 million from Greylock—led by partner Jerry Chen—alongside Amplify Partners, BoxGroup, and SV Angel. The cap table reads like a who's who of enterprise infrastructure: Reid Hoffman, Datadog co-founders Olivier Pomel and Alexis Lê-Quôc, and executives from Anthropic, Google Cloud, and Slack all participated as angels.
But the real story isn't the funding round. It's the problem Niteshift is betting will define the next phase of AI development tools, according to the company and its investors.
Building for a World That Might Not Want to Choose
Niteshift's founders, Sajid Mehmood and Conor Branagan, are pitching what they call a "full-stack cloud for coding agents"—infrastructure that allows developers to swap between AI models without rebuilding their entire environment. Think of it as avoiding the equivalent of rewriting your application every time you want to try a different cloud provider.
The duo knows infrastructure lock-in intimately. Both were early engineers at Datadog, the observability giant where Branagan became the top code contributor to the original monorepo and Mehmood led teams that shipped more than a dozen products. That pedigree helps explain why Datadog's co-founders opened their checkbooks for the round.
Rather than reselling tokens—the typical model for AI infrastructure startups—Niteshift charges on a per-minute usage basis. Customers bring their own model tokens and can switch between agents like Claude Code, Codex, OpenCode, and Pi while maintaining consistent environments, tooling, and policies. The pricing runs lean: a free tier with $10 in monthly credits, then $50 for individuals or $250 for teams. Active agent time costs 0.1 credits per minute.
It's an approach that seems almost quaint until you consider the context.
When the President Bans Your AI Provider
Earlier this year, a DeSantis administration directive briefly upended the AI vendor landscape. On February 27, federal agencies were ordered to stop using Anthropic technology over supply chain concerns. A preliminary injunction restored access in late March, but the damage was done—at least to the illusion that betting on a single AI provider carried minimal risk.
The episode, perhaps more than any technical argument, crystallized what Mehmood and Branagan had been preaching: vendor lock-in in AI isn't just inconvenient. It can be existential.
Greylock's Chen framed the Niteshift investment around "unbundling the agent from the infrastructure," according to the firm's announcement—a bet that organizations will increasingly demand separation between their choice of AI model and the development environment underneath. Whether that thesis holds depends partly on how quickly enterprises adopt coding agents at all, a question mark that still hovers over much of this market.
The Technical Scaffolding
What does Niteshift actually do? The platform reads a repository's documentation and continuous integration configuration, then automatically provisions development environments—services, databases, message queues, seeded data, even live preview URLs. Coding agents run inside these isolated cloud environments with what the company describes as unlimited concurrency.
Integrations span the usual suspects: GitHub, Slack, and Linear for triggering tasks; Sentry, Datadog, AWS, and Vercel for connecting to existing workflows. Recent additions include support for Claude's "dynamic workflows," AWS IAM role integration for least-privilege sessions, and automations that fire off webhooks or monitoring alerts.
It's infrastructure that feels almost invisible until you need it—the hallmark of good developer tooling.
Early customer names like Standard Bots, Listen Labs, and Elicit appeared in launch materials, though commercial terms remain undisclosed. The company's headcount sits somewhere in the 2-10 range according to LinkedIn data from recent weeks, a figure that will almost certainly climb now that the funding has closed.
Competing in a Billion-Dollar Frenzy
Niteshift is entering a market that has attracted staggering amounts of capital in a compressed timeframe. Cognition, the maker of Devin, reportedly raised over $1 billion at a $26 billion post-money valuation in late May. Days before Niteshift's announcement, OpenRouter—an AI gateway that routes across more than 400 models—closed a $113 million Series B led by CapitalG at roughly $1.3 billion post-money. In April, reports surfaced that SpaceX was exploring a deal to acquire or invest in Cursor, though those discussions remained unconfirmed.
Against that backdrop, a $7 million seed round might seem modest. But Niteshift isn't playing the same game as Cognition or Cursor. Those companies are building end-user AI coding experiences. Niteshift is building the plumbing beneath them.
The question is whether developers will pay for that plumbing separately—or whether the coding agent providers will bundle it so tightly that infrastructure becomes invisible, and perhaps irrelevant.
No valuation was disclosed for Niteshift's round, a detail that leaves open the question of how much dilution the founders took to bring in this roster of investors. For a team of two Datadog veterans with a live product and a thesis grounded in hard-won infrastructure lessons, the terms were likely founder-friendly. But in a market moving this fast, friendly terms and strategic position are only the beginning.
The real test comes when customers decide whether model flexibility is worth the operational overhead of managing yet another layer of infrastructure. Niteshift is betting the answer is yes. By the end of the year, we'll start to see whether they're right.
