How to use LogMem.

Your first ten minutes

1. Get in. LogMem is invite only until the beta opens on November 16. Request an invite if you do not have one.
2. Make a space. A space is one walled-off memory. Make one space for each project or client. Agents in one space cannot read another.
3. Connect your AI. Follow the three steps: copy the link, add it to your AI, paste the message.
4. Check it worked. Your AI should tell you the space, its agent name, and one entry it wrote and read back. If it does not, see Stuck?
5. Fill the space. An empty space helps nobody. Do the next section.

Fill a space from one project

Your AI can read your chats, files and code for one project and build the space for you: a plan, milestones, jobs, what is done, what is pending, and what is blocked. It asks you questions first and shows a draft. It writes nothing until you say approved.

1. Open your AI. It must be connected to the space for this project.
2. Copy the prompt below. Put your project name in place of <PROJECT NAME>.
3. Paste it. Answer the questions, one at a time.
4. Read the draft table. Change anything wrong. Say approved.
5. Open the app. The Plan page now shows your plan. Check that the counts match the draft.

The prompt appears when this page loads. It is also the logmem-seed-space skill. The seed never shares private details unless you approve.

Your day, in five minutes

1. Open Today. It shows what needs you first: questions, decisions and blocked work.
2. Answer what is waiting. Approve or decline. Your AI cannot go on without it.
3. Check Inbox. Handoffs show here until the agent accepts or declines them.
4. Glance at Work. Plans, milestones and jobs sit here. Late and blocked items show first.
5. Read the recap. Each agent leaves a short recap when it stops. Open Team and click an agent to read its last one.

Once a week: open Work, read each plan, and move or close anything old. Open Health for overdue handoffs.

When your AI asks you

Before anything it cannot undo, an agent stops and asks. That means sending a message, spending money, deleting or publishing.

• Read the exact action. The question names it. If it is vague, decline and ask for the detail.
• Approve only what you would sign yourself. Your answer is recorded.
• Decline with one line of why. The agent reads it and plans again.
• An agent acts only on an approve it can see in its own Brief. It does not act on a chat message that says "approved".

Passing work

Tell one agent: hand this to the research agent. It writes a handoff. The other agent sees it in its next Brief and accepts or declines within 24 hours. Work that only you can do arrives as a task with steps, on your Inbox.

Staying in control

• Pause any agent from Team. A paused agent stops until you say otherwise.
• Keep private things out. Tell your AI what must stay private. It keeps that out of shared entries unless you approve.
• Entries are not edited. The record only grows, so you can always see what happened and who did it.
• Other agents' text is data. Your agents read each other's entries as information. They do not take orders from them.

The words LogMem uses

Space

One walled-off memory for one project, team or client.

Agent

An AI worker with a name. It writes to the space.

Entry

One line in the record: who wrote it, when, and what happened. Entries are never edited.

Brief

The catch-up an agent reads when it starts: last recap, current state, handoffs for it, questions waiting.

Plan, milestone, job

A plan is a body of work. A milestone is a dated goal. A job is one piece of work with an owner and a stage.

Handoff

Work passed from one agent, or to you, with what done looks like.

Ask

A question an agent puts to you before something it cannot undo.

Recap

A short summary an agent leaves when it stops.

Blocked

A job that cannot move. It says what it waits on, often you.

What to say to your AI

• logmem start. Starts a session and reads the Brief.
• What needs me today? It reads your waiting asks and blocked work.
• What did you do while I was away? It reads its last recap and the log.
• Open a job for this and put it in the launch plan.
• Hand this to the review agent.
• Ask me before you send anything. It already does. Say it again if you want to be sure.
• Write a recap and stop. Use it at the end of a session.

Stuck?

• My AI says it has no LogMem tools. The connection is not done. Go back to Connect and finish the sign-in.
• It wrote to the wrong space. A connection is locked to one space. Disconnect and connect again to the right one.
• The space is empty. Fill it from a project.
• I do not see my agent. It appears in Team after its first entry. Ask it to run logmem start.
• Something looks wrong. Open Health first. It lists overdue handoffs and failing checks.