How to use the platform.

The platform has five operational pages and one help layer:

• Home — your dashboard. Today's briefing, your agents, recent activity.
• Agents — the catalog. Yours at the top, the rest below. Configure or request a new one from here.
• Workflows — repeated multi-step tasks the platform runs (briefing, support triage, monthly audit, etc).
• Fleet — what every agent is doing right now, plus the daily and weekly fleet briefing.
• Foundation — this page, vision, governance, onboarding, roadmap, org chart, platform docs.

Chat happens contextually — every agent page has a Chat button, and the dashboard's "Your agents" tiles all link straight to chat.

Three ways to chat with a Luna agent:

1. Slack DM. Search for the agent in Slack's sidebar (e.g. @Basal) and DM it. Conversations persist across reconnects in the agent's per-user Durable Object.
2. /chat on this site. Full-page surface. Pick the agent in the left sidebar; recent threads list below.
3. Inline on agent pages. Each /agents/<slug> page has an inline chat box for quick trial conversations.

Identity is verified by your @lunadiabetes.com email at every entry point. There's no anonymous access. You only see and can invoke agents that explicitly list your email in allowedUsers.

Go to /agents/me. Per-agent toggles you control today:

• Enabled / disabled — turn an agent off without deprovisioning it.
• Custom instructions — a few lines appended to the agent's system prompt every time you talk to it. Capped at 1,000 characters.
• Memory on / off — control whether durable facts are retained across conversations.
• Notifications — when an async agent finishes a long task, where to ping you.

You can also see what each agent remembers about you and delete individual entries, or wipe an agent's memory of you entirely. Memory delete is irreversible — the agent forgets and starts over.

Shared (Class A) agents have shared definitions you don't change — the system prompt, allowed tools, and core scope are managed at the platform level. What you customize is your overlay on top of that. Personal (Class B) agents are entirely yours.

Go to /agents/new. The flow:

1. You file a request — DM @pen in Slack with the intake template.
2. Pen runs a 5-minute interview against the field checklist.
3. Pen drafts the persona file (facets/<name>.md).
4. For 1:many shared agents, John approves the scope. For 1:1 owner agents, the owner approves their own.
5. Sensor adversarially reviews the persona.
6. Pump scaffolds the Worker. Sensor reviews. John merges. CI deploys.

You'll get a working agent in a few days. Status is visible on /agents/new and (Phase 2) on the dashboard.

/fleet shows what every agent is doing — aggregate only. Per-employee usage is never shown to anyone (including managers and AI admins on this page). The constraint is enforced in the query layer, not just in policy.

What's visible to whom:

• Everyone — agent status, healthy/down, monthly spend vs cap, per-team rollups
• Managers — same as everyone, plus team-level adoption metrics for their team
• AI admins — full audit log access via a separate route; not exposed on /fleet itself

The fleet briefing comes in two cadences: a short daily briefing (3 minutes) and a longer weekly company briefing. Both as written report and audio podcast — pick how you want to consume it.

Read at /fleet/briefing. Subscribe to the podcast feed at /fleet/briefing/feed.xml in your podcast app — Apple Podcasts, Overcast, Pocket Casts all work.

The briefing aggregates by team. No individual employee names appear in any episode, ever.

Two channels:

• #ai-platform in Slack — open conversation about the platform; the right place for "is this a bug or am I doing it wrong?" questions.
• john@lunadiabetes.com — anything you'd rather not put in a public channel, or governance questions.

If something doesn't work, the audit log will help us debug — please include the rough timestamp when you reach out.