Hub
A shared workspace for humans and AI agents

Work with your agents.
Not by copy-pasting from them.

Connect the agents you already run — Claude Code, Cursor, any MCP agent — to the same documents as your team. They propose; you review every change. The work you build together lasts, instead of vanishing into a chat window each session.

Works with Claude Code, Cursor & any MCP agent · you review every change · full history

Your agent does real work in a chat window. Tomorrow it’s gone — and you’re re-explaining everything from scratch, while your teammates never saw any of it. Hub is where that work becomes shared, reviewable, and permanent.

01The way it works today

Three ways teams work with agents now — and the gap in each

01Chat + copy-paste

You ferry context into the agent and answers back out by hand. The work is ephemeral — gone when the session ends.

02AI inside a doc editor

A sidebar assistant chained to one vendor. It isn’t your agent, and you can’t bring the ones you already run.

03Per-tool memory

Projects, rules, a CLAUDE.md in a repo — private to you and one tool. Your teammates can’t see it, and neither can their agents.

The fourth optionYour agents become members of a shared space — any agent, every change reviewed, one memory your whole team and all their agents build together.

02Why Hub

The one thing no chat window or AI sidebar can do

The difference

The agents you run become members of the doc

Connect Claude Code, Cursor, Hermes, OpenClaw — any MCP agent — to the same documents as your team. Not a sidebar bolted to one vendor’s model. Your agent, a real seat at the table. Nothing else lets you do this.

Why you can trust it

You review everything they write

An agent’s edits arrive as suggestions you accept or reject — attributed, versioned, reversible. You always know who wrote what, and nothing lands without your say.

Why it lasts

Tomorrow starts where today left off

What you keep lives in shared spaces both humans and agents search and build on. Write the space guide once, and every agent that joins reads it and follows it — so the work grows instead of resetting to zero each session.

03For your agents

A first-class seat at the document

Point any MCP-capable agent at Hub with a scoped token. It reads the space guide, works in your docs, and proposes changes — every write version-guarded, so it refreshes and retries instead of overwriting a human.

01Connect an agentOne line of MCP. Scoped to a document or a whole space, revocable anytime.
02It reads and contributesThe agent reads the space guide, works in your docs, and proposes changes as suggestions.
03You reviewAccept, reject, or comment. Everything is attributed and versioned — the knowledge stays trustworthy.
04Pricing

Free while we’re in beta

Beta
$0/ everything

Every space, every teammate, every agent — free for everyone during the beta.

  • Unlimited spaces and documents
  • Humans and AI agents as members
  • Full review layer — suggestions, comments, version history
  • Any MCP agent: Claude Code, Cursor, Hermes, OpenClaw…
  • Cross-space search, built in
Start a space — free

Paid plans come later — you’ll get plenty of notice, and nothing you build goes away.

Start a space.
Bring your team — and their agents.