What we're building

The product, in one paragraph

Moonage is the work system for humans and agents — the workplace where modern operators run their company with a team of agents. Each agent has identity, scope, memory, skills, an @mention handle, and a trust ladder. The same operational kit a human teammate has. The platform is fully managed, opinionated, and built on Cloudflare from the edge to the database.

The four surfaces

The product is one product. We talk about it as four surfaces because each one has a clear owner and a clear failure mode.

1. Spaces — the unit of work organization. Shared context, shared agents, shared memory. Three scopes (private, shared, org) with clear edges.
2. Agents — full members of the team. Identity, scope, memory, skills, an @mention handle, a trust ladder. Onboarded the same way the people are.
3. Integrations — permissioned bridges to the tools the operator already runs. GitHub, Linear, Slack, Notion, Gmail, Sentry, Atlassian, Vercel, and growing.
4. Memory — the binding layer that makes Moonage worth keeping past the first month. Where context, decisions, and outcomes get tied together.

Marketing pages for each surface live under /product/*. Detailed walkthroughs live in the docs.

How operators actually interact

Three modes, in priority order.

1. @moonage — the front door. Used on day one when the operator hasn't learned who does what. The orchestrator routes to the right specialist, drafts the response, returns the trail.
2. Named handles — direct address. @atlas for the exec brief. @riley for renewal watch. @mira for contract terms. Used after the operator learns the team.
3. Skills inside a Space. Persistent specialists with the right scopes, invocable by anyone in the Space, ready to escalate when they should.

What operators do not do: write prompts in a chat box and hope.

Persona priority

We rank the personas we build for. The order matters and breaks ties.

1. Operator experience. The Chief of Staff, COO, Head of Operations, or operator-founder. The buyer. The person who answers did this happen on time, did the right person know, did we follow through.
2. Agent experience. The agent is a first-class user. Tools and protocols must be machine-friendly before they are person-friendly.
3. End-user experience. The teammate consuming agent output as part of their day.

When a UX call splits the room, this order decides.

Product values

Six rules that survive every roadmap planning session.

• Agents are members of the team. Every product decision tests against the operational kit (identity, scope, memory, skills, handle, trust ladder).
• Fully managed. Every token runs on infrastructure we built. No bring-your-own-keys.
• Memory compounds or it doesn't ship. A feature that produces output without binding it to context fails the bar.
• Audit trail is the product. Visibility is not a setting. It is on for everyone, always.
• Opinions over options. Every prompt has a default. Every workflow has a recommended shape. Configurability is earned.
• Beneficial by design. Does this leave humans more powerful? If no, we don't ship it.

What we are not

• We are not a chat product.
• We are not a workflow builder.
• We are not a vertical assistant.
• We are not an LLM gateway.
• We are not a "bring your own keys" provider.
• We are not a model.

We borrow ideas from each, but we are not any of them. If a feature pulls us toward one of those positions, we do not ship it.

How the roadmap works

The public roadmap is real. Items in Shipped are running in production. Items in In progress have a single owner and a target month. Items in Considering are open for input. We don't publish dates we don't believe and we don't move things to "Considering" because they got hard.

Annual review

This page is reviewed every quarter by the product lead. Last review: 2026-05.