Common questions.

Drop your files into the 01_Inbox Drive folder. Then send an email to claw01@hive.au with a short message. Something like: "Hey, it's Sam. I've added a brief for the Acme campaign. Come back with questions or a plan." That is all it needs to get started.

The email watcher checks every 2 minutes, so it will pick it up quickly. See the full process map.

Everything lives in the shared Drive under HiveAU. The five folders are: 01_Inbox for new jobs, 02_Active for work in progress, 03_Output for finished deliverables, 04_Archive for completed work, and 00_System for agent config and memory.

You drop files into 01_Inbox. The main agent moves them to 02_Active when the job starts, and puts the outputs in 03_Output when done.

Wayne adds you. He shares the inbox Drive folder with your Google account and adds your email as an approved sender. Once that is done, the flow is the same as everyone else. Drop files, send the email, wait for a response.

Check the task folder in 02_Active. Each job gets its own folder, and inside that folder is a LOG file updated at every step. It records when the job was read, what the agent decided, which specialist ran it, and what was produced.

If the LOG file shows the job is still running, it is still running. If it says done, check 03_Output for the deliverable.

Just ask it via email or Slack. HiveXS searches a local knowledge base built from everything in the workspace folders. It can find past briefs, outputs, and reference material from completed jobs that have been archived.

The more the workspace accumulates, the better the knowledge base gets. Completed work in 04_Archive is still searchable.

A few key differences. ChatGPT knows nothing about Hive, your clients, or your past work. HiveXS is built into the workspace, so it knows the files, the history, and the way we work. It also acts, not just answers. It reads emails, moves files, routes tasks, and writes outputs. It is a system, not a chat window.

It also has approvals baked in. Everything sensitive goes through a person first. ChatGPT does not have that.

Yes. Post in #hivexs-agents. It reads that channel. Good for quick questions, status checks, or short requests that do not need a full inbox job. For anything with attached files or a proper brief, email is the cleaner path.

It does not replace anyone. Strategy, voice, judgement, and the final call on quality are all still human. It does not talk to clients directly. It does not send anything externally without going through a person first. It does not act without approval on anything sensitive. It does not expand its own access or spin up new agents without Wayne's say-so.

The team does the part that matters most. HiveXS handles the parts that drain time without adding much thinking.

Every job has a log. If something went wrong, the log shows where and why. Email Wayne or post in Slack with the task name and what happened. Nothing is deleted without being recoverable first, so there is always a trail to follow.

If a piece of output is wrong, just flag it the same way you would flag any other work. Say what was wrong, and what you needed instead. The system learns from correction.

It is a local SQLite database on the Mac that runs HiveXS. It indexes every file in the workspace folders using BGE-small embeddings, and uses a Llama 3.2 re-ranker to surface the most relevant results when something is searched.

A nightly review job keeps it current with any new or changed files. You do not need to do anything to maintain it. See the system map for how it fits into the architecture.

Every API call is logged in a local cost ledger. A nightly reconciler tallies usage and emails Wayne a summary. If you want to see the running total or drill into a specific job, ask HiveXS directly and it can pull the figures from the ledger.

Talk to Wayne. New agents are scoped and set up deliberately, not automatically. There is a roadmap: a writer agent and a researcher agent are planned for later phases. If there is a repeating task that would benefit from its own specialist, that is worth raising. The system is designed to grow, just with intention.