VIVUU · The AI Work Terminal
What needs your attention? The AI agent control plane that brings you only the decisions.
Not another chat box. VIVUU's agents watch your connected systems, do the routine work, and collect the evidence — you monitor and approve from your phone, and make only the decisions that require judgment.
vivuu · developer
What needs your attention?
Checkout failure rate increased 4.1×
- impact
- ~18 failed payments / hour
- cause
- Regression in deploy #847
- done
- patched → 214 tests green → staging ✓
Flaky test: fixing cold-cache race in CI
portrait by design — the shape of the terminal to come
Why a terminal, not a chatbot
Built on three rules
Attention first
The home screen is never an empty prompt. It's a ranked queue of decision-ready items: what changed, why it matters, what was already done, and the one decision you owe.
Evidence before trust
No claim without proof one tap away — diffs, test runs, staging logs, metric trends, each with a checksum. You approve what you can inspect, not what a model summarized.
Deliberate approvals
Consequential actions are gated by policy, framed full-screen, and confirmed by press-and-hold. Denials require a written reason. Everything lands in an append-only audit trail.
The loop
Agents run the loop. You own one step.
VIVUU works continuously through eight stages. Seven belong to the agents. The approval — the one that changes production, sends the message, moves the money — is deliberately yours.
Role packs
Starts where the work is measurable
A role pack tunes the attention queue, the connected systems and the approval policy to the job.
role pack
Developer
Broken builds reproduced and bisected, dependency advisories patched inside SLA, flaky tests diagnosed, PRs pre-reviewed. Agents work in isolated sandboxes; pushes and pull requests wait for your approval.
role pack
Founder
Revenue anomalies, churn-risk signals, search-performance drops, metric drift: the same loop pointed at the business instead of the codebase.
role pack
CEO
Decisions only: high-risk approvals, KPI drops and one operations pulse. Ask an executive question; delegate the rest.
role pack
Basic
Your own Google or Microsoft account: find a contact, find a file in your Drive or OneDrive, and send it, summarized or translated. Nothing is emailed until you approve it.
Guides & field notes
How to run agents with a human at the right step
- Guides · Definitions
What Is an AI Agent Control Plane?
A plain definition of the AI agent control plane: policy, approvals, evidence and audit between agents and consequences, and how it differs from observability.
- Guides · Agent governance
Human-in-the-Loop Approvals for AI Agents
How human-in-the-loop approval works for AI agents: when an agent should stop and ask, what a reviewable approval contains, and policy tiers from ALLOW to DENY.
- Guides · Agent governance
AI Agent Guardrails: The Four Layers That Matter
The four layers that actually constrain an AI agent (prompt, tool, policy and gate), why prompt-level rules are the weakest, and concrete examples of each.
- Guides · Agent operations
Monitor Claude Code from Your Phone
How to watch and approve Claude Code and other background coding agents from your phone: the state you need, today's workarounds, and a purpose-built terminal.
- Field notes · Running agents
AI Agent Failure Modes: 6 Seen in Production
Six failure modes from running coding agents on real repos behind an approval gate: confident wrong plans, empty successes, vocabulary drift, lost workspaces.
- Field notes · Running agents
What AI Coding Agents Actually Cost Per Task
Measured cost of a gated coding agent end to end: about ten cents a task and two model calls, and why the model bill is cheaper than review time and re-runs.
Questions about VIVUU
- What is VIVUU?
- VIVUU is an AI agent control plane, built as the AI Work Terminal. Agents watch your connected systems, do the routine work and collect the evidence; you get a ranked queue of the decisions that need judgment, and every action lands in an append-only audit trail.
- How is VIVUU different from an AI chatbot or an agent dashboard?
- A chatbot waits for a prompt and a dashboard shows activity. VIVUU opens on what needs a decision: each item says what changed, why it matters, what was already done and the one decision you owe, with the diffs, test runs and logs one tap away.
- What does an AI agent control plane do?
- It sits between agents and consequences. Every requested action is checked against policy before it runs: routine work is allowed, consequential actions wait for a human approval, moving money needs two approvals, and changing a security control is denied outright. Anything not explicitly permitted is refused.
- Which agent actions need my approval?
- Anything that leaves the sandbox: pushing a branch or opening a pull request, deploying to production, sending a message, or moving money. Reading connected records, drafting and editing inside the sandbox run without asking.
- Can I monitor and approve AI agents from my phone?
- Yes. The terminal is a portrait, installable web app designed for the phone. Approvals open full-screen with the evidence attached, are confirmed with press-and-hold, and a denial requires a written reason.
- Is VIVUU free?
- The terminal interface is open to explore with no sign-up, and it stays empty until it is connected to your own systems. Connecting is currently limited to invited accounts, and there are no public paid plans yet.
- Who is VIVUU for?
- Role packs tune the queue to the job: Developer (CI, deploys, dependencies, flaky tests and PR review), Founder (revenue signals, churn risk and search-performance anomalies), CEO (high-risk approvals, KPI drops and one operations pulse) and Basic (find contacts and files in your Google or Microsoft account, and send a file, a summary or a translation once you approve it).
Open the terminal. See what needs you.
Look around the interface: the queue, approvals, evidence and settings. It fills with real cards once it is connected to your systems.