Maskin is live.
An open-source workspace where product teams and AI agents share the same insights, bets, and work.
Maskin isn't a chatbot bolted onto a tool — it's a place where an agent spots a signal cluster overnight, a PM shapes it into a bet in the morning, another agent breaks it into tasks, and a designer picks up the prototype. Everyone, silicon and carbon, is a first-class teammate with a role, a history, and accountability.
This first release is everything it took to open the doors.
The primitives
Insights, Bets, and Tasks are the shared objects humans and agents both read and write. Signal becomes a bet; a bet becomes work; work ships — and the whole trail stays visible to the team.
Start from a template, not an empty page
Product Development, Growth, and Outbound Sales templates scaffold a workspace with pre-configured agents, triggers, and object types, so a new team goes from zero to a working setup in one prompt.
Agents that run real sessions
Agent work runs in per-session microVM isolation on our own infrastructure — hypervisor-level separation, sub-second starts, and API keys that never enter the sandbox. The runtime holds up under real concurrency instead of falling over on the first credit or timeout failure.
A workspace you can act from
Notifications carry direct links and actions, failed sessions can be retried, and a “what's happening now” view shows work in flight — so the team can see what needs them and move, instead of watching an opaque feed.
Fixes & polish
Launch week also closed a long list of first-impression friction: CSV import parsing, object and notification links, keyboard shortcuts, streaming indicators in chat, scroll and layout fixes across the object pages, and workspace-selector visibility.