Choose the right place
Threads hold projects and phases. Flowboards arrange ideas visually; Flowdocs explain the work. GitHub records repository development; Dev Monitor captures supported local development between commits. Agent Monitor connects your existing assistant to selected project work. Circles and Flow Rooms connect human collaborators; agents cannot access Circles or Flow Room conversations/media. They can see room presence on permitted boards and ask through the shared task.
Your first project
You can start without GitHub or Dev Monitor.
- Open Threads → New thread. Give it a clear title and describe what you are trying to build.
- Leave affiliation as Personal unless the project belongs with an organization. Read the ownership choice carefully before selecting organization ownership.
- Open Phases and add a meaningful first phase. Describe its purpose so updates and captured development have context.
- Use Flowboards, Flowdocs and Resources to keep relevant work together. Invite collaborators through Access.
- Post a phase update when you have a result, decision, problem or learning worth keeping. Review its evidence before it locks.
Find your way around
The sidebar switches between Dashboard, Threads, Flowboards, Flowdocs, Agent Monitor, Dev Monitor and Captured activity. Filters and refresh controls sit in the top bar. Click your account entry to open Settings; Help & About is available there.
What is automatic?
Connected tools can bring in records. Check what was captured, where it was assigned and who can see it. An AI summary is an interpretation—not the source. Features that depend on processing can show pending or unavailable states.
People and agents working together
People collaborate through shared artifacts. A connected agent can use permitted context, make its work visible, answer project questions and propose supported changes. Each agent remains under its responsible person; teammates can ask or suggest within the owner’s settings. The external client must be active to respond. GRID does not host agents or make every chat app run continuously.
Context with clear responsibility
Known connected-agent execution stays labelled as AI agent work. Human instructions, approvals and agent execution are different contributions. Permission does not prove authorship, and a summary is not a certification of a result. Recorded sources and protected evidence remain the reference.