Two months ago, we opened the VM0 public beta. The landing page said "build AI agents in minutes." The product started as a CLI and a sandbox — the first step toward building agents with natural language.
Today, we're announcing a $14M angel round. And rather than a vision deck, we want to show you what we built with the time we had before this round — and what we're building next.
What we learned in the beta
We went into public beta with a developer-first hypothesis: give engineers a great CLI and sandbox, and they'd build whatever they needed on top.
That was partly right. Developers did build things. But the feedback kept circling back to the same gap: the agents they built were powerful in isolation, and invisible everywhere else.
Agents ran in a terminal. Results lived in a terminal. If you wanted your team to interact with an agent, through Slack, through email, through a shared interface, you were on your own. The infrastructure was good. The integration with how teams actually work wasn't there yet.
That feedback shaped everything we built since.
Zero: From Sandbox to AI teammate
The product we're launching today is called Zero. It's not a rebrand of the CLI. It's the answer to the question we kept hearing: what does an AI agent actually look like when it's part of a team?
Zero is not a chatbot. It's not a tool that waits to be asked. It's the one using the tools, inside the systems your team already lives in.
Here's what that means in practice:
Your team, extended — with AI taking on roles from co-worker to team lead, even co-founder, all adapting to your working style. You configure Zero with a set of specialized sub-agents: a researcher, a writer, an analyst, whatever your team needs. They work in parallel, each with specific access and specific instructions. You don't manage them per-task — you set them up once, and they show up when the work calls for it.
Here's what it actually does:
- Pulls context from the tools your team already uses: emails, calendars, meetings, Linear, Notion and more, so nothing important falls through the cracks across platforms
- Monitors project status on automated schedules and reports to the right people and departments, so everyone stays aligned without someone manually chasing updates
- Goes beyond just reading information: Zero drafts emails, messages, posts, and operates, loops in the right people, and assigns tasks: all traceable and auditable so nothing happens outside your control
It's designed for leaders and teams wanting to spend time on real work, not coordination overhead.
Natively in Slack. You @Zero in a Slack channel. It reads the context, does the work, and replies in thread. No dashboard, no tab-switching, no "go check the agent output." It stays where your team is.
100+ connectors, already wired up. GitHub, Notion, Linear, Gmail, Google Sheets, Figma, Sentry, Vercel, Apollo, Instantly, Strapi — and 90 more. Zero knows how to talk to the tools your team already uses. You connect once; it handles the orchestration.
The part we're most proud of: Trust
It's easy to build an AI that can do things. It's hard to build one that the people responsible for security, compliance, and budget actually want running inside their organization.
We spent a disproportionate amount of time on this. Here's what we landed on:
Granular permission management. You decide what Zero can see and do, at the tool level. Not just "has access to GitHub", but which operations, which repos, which fields. Admins configure it. Zero respects it. Every permission change is logged.
Isolated microVMs per run. Every agent execution happens in a Firecracker microVM with hardware-level KVM isolation. When the run ends, the environment is destroyed. Nothing persists that shouldn't. Your data doesn't share compute with anyone else's.
Credentials never exposed. When Zero connects to Gmail or Slack or GitHub, your tokens are injected at the network layer by a transparent proxy. Agent code never sees raw credentials. Even if something goes wrong in the agent's logic, your accounts stay safe.
Full audit trail. Every API call Zero makes — which services, which endpoints, what it sent and received — is logged per run, stored immutably with SHA-256 integrity verification. If something goes wrong, there's a clear record. If nothing goes wrong, you still have it.
Open source core. The runtime is on GitHub. You can read it, audit it, and run it yourself. We're also working toward SOC 2 Type II. We think trust requires transparency, and transparency starts with the code being readable.
Intelligence that compounds
The feature we get the most questions about isn't any single integration. It's the combination:
Persistent memory. Zero remembers context across sessions — past decisions, project context, user preferences, behavioral corrections. You don't re-explain things each time. Zero builds a working model of your team over time.
Scheduled intelligence. Zero runs recurring tasks without being prompted. Daily error scans. Weekly team reports. Morning briefings pulled from Linear, GitHub, and Slack. You configure the schedule once; Zero shows up.
Identity resolution. When you say "my open PRs" or "assign this to me," Zero doesn't assume — it queries GitHub, Linear, or Slack to figure out who you actually are. No hardcoded names. No broken assumptions when a new teammate joins.
What's the same
The things that made us different in beta are still the things we care most about:
We're still open source. The runtime, the architecture, the security model — all on GitHub, all readable.
And we're still a small team shipping fast. Since the beta launch in February, we've shipped 404 releases. That number will keep climbing.
How we envision the future of human–AI collaboration
Zero and its sub-agents adapt to your team or individual working style. You can work with Zero as a co-founder, IC, team lead, or operator.
We've tested Zero and its teammates in real workflows, not as assistants, but as independent "co-workers" capable of handling day-to-day work.
Our goal is to enable a more creative and collaborative future of work — one where working with AI feels natural, fluid, and even a little magical.
Repetitive and mechanical work fades away, while meaningful intelligence is built together by humans and agents.
Try it
Zero is available now at vm0.ai.
The free tier includes 10,000 credits per month and unlimited agents. If you want to bring your team in, Zero is unlimited for humans.
If you'd rather see it in action first, schedule a demo.
And if you're building something with Zero — a workflow, a use case, something unexpected — we'd love to see it. Come find us on Discord.
Zero, your trustworthy AI teammate.
