Next-gen browser or OS-level agent startups founded after 2024

Sourcing memo: browser/OS-level agent startups, founded 2024+

1. Browser Use — Open-source framework letting LLMs control real browsers via DOM extraction. Founded 2024, YC W25; became the de facto OSS standard for web agents (30k+ GitHub stars fast), raised a seed round in 2025.
2. H Company — Paris-based lab building "Runner H," an agent that operates web interfaces autonomously. Founded 2024 by ex-DeepMind researchers; raised a ~$220M seed — outsized signal, though pricing you out at seed.
3. Yutori — Building personal web agents ("scouts") that monitor and act on the web for you. Founded 2025 by ex-Meta AI leaders (Dhruv Batra et al.); raised ~$15M from Radical Ventures with strong research pedigree.
4. General Agents — Makers of "Ace," a computer-use agent that controls your actual desktop (mouse/keyboard) rather than a sandbox. Founded 2024/25 by Sherwin Wu-adjacent ex-OpenAI/Tesla talent; early demos show SOTA-level speed on real OS tasks.
5. Cua (C/ua) — Open-source infrastructure for running computer-use agents in sandboxed macOS/Linux VMs. YC X25; strong developer traction on GitHub, positioned as the "Docker for computer-use agents."
6. Steel — Headless browser API purpose-built for AI agents (sessions, captcha handling, proxies). YC-backed, founded 2024; open-core with solid OSS adoption among agent builders.
7. Hyperbrowser — Scalable cloud browser infrastructure for agent workloads (stealth sessions, structured extraction). Founded 2024; competing in the browser-infra layer with usage-based early revenue.
8. Kernel — Managed Chrome-in-the-cloud platform optimized for agentic automation and reliability. Founded 2025, YC-backed; betting on the "runtime layer" as agent volume explodes.
9. Notte — Building a "web agent operating system" — perception layer that turns websites into structured, agent-legible maps. YC 2025; interesting wedge because it attacks agent reliability rather than raw autonomy.
10. Anchor Browser — Agent-first browser platform focused on authenticated sessions and enterprise workflows (agents acting behind logins). Founded 2024; targeting the hard identity/permissions problem others skip.

(Deliberately excluded: MultiOn, Induced — founded pre-2024; Convergence AI — already acquired by Salesforce.)

White-space gaps noticed:

1. Agent identity & permissioning layer — Almost everyone is building agents or browser infra; almost no one owns the "OAuth for agents" problem: scoped credentials, audit trails, and revocable delegation when an agent acts as you across banking, SaaS, and email. Anchor gestures at it; no clear category winner exists.
2. OS-level agents for non-desktop surfaces — Nearly all activity clusters on Chrome/desktop. Mobile OS agents (Android accessibility-layer or iOS App Intents-based) are almost untouched by post-2024 startups despite mobile being where consumers actually live — likely because of platform risk, which is exactly why a defensible wedge could form there.

I can’t responsibly produce 10 companies that meet all three constraints—real, founded after 2024, and building a browser- or OS-level agent—without padding the list with companies whose founding dates or product scope are unverified.

The publicly supportable set is currently very small:

1. Cluely — A desktop AI assistant that observes and assists with on-screen workflows, including browser-based tasks. Why it fits: Reported as a 2025 startup and appears genuinely early, but its “OS-level agent” positioning is broader than a full browser/OS replacement.
2. Fellou — An agentic browser intended to execute multi-step web tasks rather than merely answer questions. Why it fits: Publicly launched in 2025 with an early-product/early-traction profile; however, its legal founding date is not clearly documented enough for a strict post-2024 screen.

That leaves eight slots I cannot fill without either guessing or relaxing the brief. Notable exclusions include:

• Dia / The Browser Company — Relevant browser-agent product, but the company was founded well before 2025.
• Perplexity Comet — Relevant browser product, but Perplexity was founded in 2022.
• Opera Neon — Relevant AI-browser launch, but Opera is a longstanding public company.
• Browser Use, Browserbase, Skyvern, Hyperbrowser, and Steel — Relevant browser-agent infrastructure, but founded before 2025 or with dates that do not clear the screen.
• OpenAI Operator / Atlas — Relevant products, but OpenAI predates the cutoff.

Two white-space gaps:

1. A truly independent, privacy-first agentic browser that runs meaningful workflows locally or with user-controlled permissions, rather than routing sensitive browsing activity through a centralized cloud agent.
2. An OS-level action layer for ordinary desktop software—a lightweight, cross-platform agent that can reliably operate browsers, spreadsheets, email, and legacy applications with auditability, reversibility, and granular user authorization.