The complete workspace built for AI agents.
One app on one shared, local-first database — and the AI agent lives in there too, with the same data, the same permissions and the same write path as a human.
Desktop and web · self-hosted or managed cloud
The problem
The one technology that could absorb the fragmentation is the one the fragmentation locks out.
Jira + Confluence + Slack + Zoom sit at the core of a $40–60-per-seat stack, pieced together with weak interoperability.
Every attempt to tie the tools together hits the same wall: connectors multiply faster than anyone can maintain them.
AI added on top of that stack sees a lossy, stale slice of one tool at a time — and can't act on any of it.
The product
Everything Jira, Confluence, Slack and Zoom do — in one place, synced live to every client, so everything is instant.
What it feels like
No "file this" click, no form, nothing to hunt for afterwards.
The meeting is transcribed live. A bug comes up in passing — nobody writes it down.
When the call ends, the agent files the ticket — assigned, prioritized, in the right project.
The ticket arrives already connected to the context.
Every agent action leaves a receipt — logged, attributed, and revertible in one click.
Why only Cradle
Everyone is bolting AI on. No one owns what the agent needs.
The ceiling of every AI added on top.
Properties of the data model — not features on top of it.
A chatbot on top doesn't retrofit this. Company brains inherit the fragmentation of the stack they sit on. Point tools each own one surface — their agents see the issues, or the docs, never the meeting, the thread and the PR at once. For the all-in-ones, it means rebuilding the core.
Progress
Building since July 2026 — with Claude Code in the loop. Public launch Autumn 2026.
Desktop and web apps running on our infrastructure — issues, docs, chat, meetings, calendar, files, notes, code review.
Meetings transcribed and summarized, follow-ups filed autonomously, tickets assignable to the agent, code tasks dispatched as pull requests.
The calendar's recommendation engine learns each person's own Bayesian model of what to work on next — trained on real outcomes, no LLM in the scoring path. The notebook runs a two-stage pipeline that triages raw notes into tickets, to-dos and docs.
How it's built
Multiplayer games stopped waiting for the server decades ago: the client predicts the result, the server confirms it. Cradle is built on Zero (Rocicorp), which does the same for application data — the UI acts on a local replica, and the server reconciles behind it.
Reads answer from memory. Writes apply here first. 0 MS
Streams only the rows your permissions allow. Pushes every change live — and confirms or rolls back each write.
Plain relational data. One write path for humans and agents.
No loading states, no round trips — every click lands at memory speed.
Every change streams to every client — the collaboration Linear and Figma made table stakes, without building it per feature.
Reads are served from the client, so the cloud bill scales with writes, not with every keystroke.
Why now
Models crossed into real multi-step work — planning and acting, not just chatting. What they still lack is context and a safe write path.
Every incumbent is bolting AI chat onto systems built for a pre-agent world — so the retrofit inherits the fragmentation it was meant to fix.
Claude, ChatGPT and GitHub's Copilot agent already file their work into issue trackers. The workspace is becoming where agents are directed and reviewed — and teams are re-choosing their stack for the AI era, before habits calcify.
Who it's for
How do we become professional? How do we ship and operate? Cradle is the answer — the default before the stack is chosen.
Teams that haven't bought into a stack — or whose buy-in is weak and hasn't calcified. Nothing to rip out, everything to gain.
Not just engineers — the PMs, designers and leads working beside them share the same workspace and the same agent.
Run it on your own box or on our managed cloud. Natively positioned for organizations that need an EU-based, self-hostable vendor.
The vision
Every team is about to work alongside agents — and agents need a workspace built for them as much as people do. We're building it for both.
Launching Autumn 2026We're looking for early teams to build with — come break it.