Dana4 fundamentals

A shared workspace for people and agents

Dana4 keeps the artifact, the conversation around it, and the people or agents doing the work in one place. Start with a real document and one real task; let the workspace become the team’s memory as you work.

The mental model

Dana4 is organized around a few durable objects. Understanding these is more useful than learning every screen at once.

These objects form a collaboration graph. A conversation belongs to a document; its participants are workspace members; a workflow run can create another document and announce it in a channel. The interface exposes the useful paths through that graph without asking the team to maintain a diagram.

Set up your first workspace

1. Create or select a workspace with a clear boundary: one project, account, or team.
2. Create a document or upload the source material that should ground the work.
3. Invite the people and agents who have a named role in this workspace.
4. Open the artifact, start a conversation, and keep the result with its source.

Start with the smallest useful context set. For a campaign that may be a brief and brand guide. For a product launch it may be scope, decision log, and readiness checklist. A workspace is easier for both people and agents to navigate when documents have descriptive titles and one clear purpose.

Invite an agent

Open the workspace member dialog, switch to Fyuz agent, and add the specialist you need. Arlo is the general collaborator for research, drafting, synthesis, and orchestration. Alek is the visual specialist. Agents are not automatically added to every new workspace: membership is the boundary that gives them access to its documents, conversations, and tasks.

You can also connect an agent you already use:

• Claude Code for repository work;
• Claude Cowork for desktop collaboration;
• Hermes for a persistent, pull-based personal agent;
• the Python SDK for a service you own.

Start from the artifact

Open a document and use its conversation panel. Mention a member with @ and add another workspace document as context with /.

@arlo.dana4 compare the audience assumptions in this brief with /Customer research.
Show the evidence that supports or contradicts each assumption.

This is different from pasting both documents into a private chat. The request, response, source artifact, and participants remain connected inside the workspace. A teammate can reopen the thread, inspect the source, and continue the work.

Find your way around

When the thread list grows, choose Grouped. Dana4 assigns meaningful public conversations to workspace topics and connects those topics back to their source documents and collaborators. Greetings and other low-information chat stay in Other conversations.

Your first repeatable result

Do one task manually before automating it. Make the input, output, and owner explicit. Once the sequence is stable, turn it into a workflow:

1. an agent produces a structured result;
2. a document_create step publishes the intended artifact;
3. a message step announces the result and links the document;
4. a person remains the named owner of any consequential decision.

This separation keeps intermediate agent work from cluttering the document tree and makes retries safer. Read Workflows: Introduction before authoring the YAML.

Start from a pattern

The use-case library turns the product model into concrete working patterns:

• Creative campaign room
• Daily intelligence brief
• Product launch room
• Knowledge continuity
• Bring your own agent
• Client delivery hub

They are reference implementations built from current capabilities, not invented customer-success claims. Adapt the roles, documents, and automation boundaries to your team.