Inkbox raises $5.3 million and launches cloud-hosted AI agents with persistent computers

Inkbox has closed a $5.3 million seed round and launched cloud-hosted AI agents that each get their own persistent computer and a set of real communication channels.

The San Francisco company came out of Y Combinator’s Summer 2026 cohort. It announced the funding on November 10, 2026, with Peak XV leading the round.

What Inkbox actually launched

The core product is an AI agent that lives on a cloud computer and stays there. Its files, installed tools and accumulated context carry over from one session to the next.

The agents also wake themselves up. When a message or task lands through a connected channel, the agent activates and gets to work without someone needing to press a button.

Those channels are the ones people already use. Inkbox says its agents integrate with email, iMessage, WhatsApp and Slack, so an agent can be reached the same way a coworker can.

Agents can also run scheduled or recurring jobs. The agents enter a sleep mode when there is nothing to do, to keep costs down.

Security is part of the pitch. Inkbox pairs its agents with secure real-world channels and a vault for storing credentials, so an agent can log in to services without passwords floating around in plain text.

Agents can sign themselves up for services through what Inkbox calls a provisional workspace process, which spares a human from walking each agent through every registration screen.

The money and the people behind it

Peak XV led the $5.3 million seed round. Participants included Guillermo Rauch, Henri Stern, Jonathan Frankle, Cristina Cordova and Dan Levine.

Inkbox was founded in 2025 by Ray Liao, Alex Wilcox and Dima Vremenko. It started as a three-person team in San Francisco.

Inkbox reached $300,000 in annual recurring revenue by September 2026, within four months of its Y Combinator launch. The company has also reportedly processed millions of messages for agents since launch.

Why the founders built it

The origin story starts with a familiar headache. Getting an AI agent to operate across multiple channels meant wrestling with logins, two-factor authentication and a patchwork of separate integrations.

The founders’ answer was a unified API layer. In plain terms, it is one connection point that lets an agent keep the same identity whether it is sending an email, replying in Slack or texting over iMessage.