AI-first team collaboration
Two kinds of intelligence.
One workspace.
Chat, whiteboards, kanban, clients, calendar and calls for small teams — with Claude as a genuine member of the team. It sees only what you allow, does real work with real tools, and everything it touches is on the record.
Free while in beta · Connects to Claude Code & Claude Desktop over MCP
Pulse blue is human
Your mentions, your selections, your actions. Everything people do wears the cool color.
You can tell what AI touched at a glance — without reading a word.
Ember is always Claude
Its messages, its cursor, the badge on anything it can read, every line of the audit log. Ember is never decoration.
Everything a small team runs on.
Nothing bolted on.
Claude is on the team
Mention @claude in any channel and it reads the project context you’ve exposed, then actually does the work — tasks, notes, client records, emails, events, whole prototypes. Every reply cites the exact sources it read.
Chat that becomes knowledge
Channels, threads, reactions — and any message promotes to a tagged, AI-readable note in one click. Conversations turn into things Claude can use.
Kanban, your workflow
Per-project boards with custom status columns, groups, priorities and #numbers. Claude files, moves and comments on cards like anyone else.
Whiteboards with an AI window
Infinite boards for the messy thinking. Mark a frame AI-visible and Claude reads it as structure — stickies, shapes, connectors — not pixels.
Clients, in the same place
A pipeline per project with stages you define, contact timelines, and email threads that link themselves to the right client by address.
Reminders that reach phones
Project calendars with wall-clock times and push reminders — 15 minutes before the meeting, on any device, tab open or not.
Email without leaving
Claude drafts and sends real email to anyone. Replies flow back into the channel and thread on the Mail page — the outside world, bridged in.
Calls with no middleman
Voice, video and screen share — browser to browser, peer-to-peer. Shares land on a zoomable full-screen stage; a call rings every phone via push. No media server ever sees or hears your call.
Charts from live data
Ask for “a bar chart of yesterday’s sales” — Claude runs the integration, gets real numbers, and renders the chart inline in its reply.
Integrations that write themselves
Ask Claude to connect your Postgres, Redis or any HTTP API. It authors the connector, posts a config form, then debugs its own code when a run fails. Secrets are encrypted and never pass through the model.
Prototypes at a link
Claude publishes click-through concepts, mockups and interactive demos as artifacts the whole team can open — and iterates on them without breaking what worked.
Claude Code, connected
One command connects the workspace to Claude Code or Claude Desktop over MCP — 26 tools with per-user tokens, so your dev environment can read and write the same projects.
On the record, by design
Nothing crosses to AI unless it’s marked readable — notes, frames and channels are opt-in. Every read and every write lands in a human-legible audit log.
Up and running in a minute
- 1Create your workspace. Name it, pick a project — that’s the whole setup.
- 2Invite the team. Claude is already in the DM list, waiting to be useful.
- 3Choose what AI can see. Flip the readable flags — then put it to work.
Ready to add a teammate?
The kind that reads everything you let it, forgets nothing, and files its own tickets.
Get started — it’s free