Good questions.
Straight answers.

Tempest is an AI Chief of Staff we're building for startup teams. The idea is to learn how your company works, help people and AI agents get the right brief, and keep you from having to chase every update. Today, the alpha starts with company knowledge and decisions. The coordination experience is still being built.

Teams where the founder is still the person everyone comes to for context, decisions and follow-ups. If you have people and AI agents doing real work, but you still have to connect all the dots yourself, that's the problem we're working on.

It's meant for the whole team, even if the founder is the first person to bring it in. The aim is for people to find the background and direction they need without asking you to explain every decision again.

Memory is where we started. Knowing why a decision was made is useful, but we want to take it further: use that background to prepare the next job, help hand it to the right person or agent, and see what came back.

No. You still set direction and make the calls that need your judgment. We're trying to reduce the repeat briefing and follow-up around those decisions, not replace you.

You can use the Company Memory, decision history, source-linked knowledge, Living Roadmap and Team Workspace foundation. These keep company context together. The new team coordination and agent workflow isn't available as a finished experience yet.

Not yet. The backend and local tests have moved ahead, but the current interface still shows the earlier foundation. The animated diagrams on this website explain where we're going. They aren't recordings of features you can use today.

The existing Company Intelligence foundation: company context, evidence, assumptions, decisions and Living Roadmap. Northline and the example workspaces use sample data, not real customer results. It is not a demonstration of automated agent execution.

Bounded briefs, user-controlled handoffs and local MCP/result-capture flows have been tested in development. Those tests do not establish that a real coding task ran, an outcome was independently verified, or every provider's desktop app opened successfully.

Automatic routing, proactive follow-through, autonomous execution and verification are not generally available production capabilities. There is no public release date promised for them here.

Not as a generally available production capability. The development-tested route is user-controlled handoff and result capture, with explicit scope and review. Opening an app is not proof that an agent has run the task.

User-controlled handoff paths have been implemented and tested locally. They are not a claim that all three desktop clients have completed real end-to-end tasks or are shipping as automatic production integrations.

No. The intended model works across existing people, tools and agents while preserving company-wide reasoning. We state current availability separately; naming a tool does not mean a production integration is available.

A clear job for someone to pick up: what to do, why it matters, what context helps and what they shouldn't change without asking. We're developing this so you don't have to rebuild the background every time you hand something over.

No. A handoff request, an opened client, a user-reported execution and an accepted outcome are different states. Tempest's product direction keeps those distinctions visible.

No. Reported or UNVERIFIED results are not the same as evidence checked or outcome accepted. A result being captured must not silently turn it into company truth or a completed roadmap item.

You and the people you've authorized. We're designing around clear scope and approvals, especially for sensitive actions. You shouldn't have to choose between getting help and losing control of your company.

We are not offering a generally available cloud MCP service here. Local MCP testing is a development step, not evidence of safe remote multi-tenant access. No public product API or cloud OAuth capability should be inferred from it.

Read the published Privacy Policy and Terms of Use, and ask us about the exact alpha workflow you intend to use before sharing sensitive company information. We do not claim blanket security certifications or risk-free access.

No. We want you to spend less effort briefing, coordinating and following up. We still need to measure that with real teams and accepted work, so we won't promise a percentage or a number of hours we can't back up.

Leave your email in the form on the homepage. We'll reach out about availability and whether the alpha fits your team. We can't promise an invitation to everyone just yet.

The current website form asks for your email address. It also presents the Terms of Use and Privacy Policy before you request access. You do not need to complete a long application questionnaire.

Of course. Use Book a demo on the homepage to pick a time. Tell us where work gets stuck in your team and what keeps landing back on your desk. That's a useful place to start.

No finalized public pricing plans are announced here. Company fees, seats and usage are being explored; previously discussed prices or package names should not be treated as live offers.

Do not assume that Tempest includes or resells a third-party agent subscription. Existing authorized tools and usage rights are separate from the product direction described here.

Email [email protected]. If you have already requested alpha access, use the same address you applied with so the team can identify your request.