Skip to content
Ryan Fong
← Back to catalogue

Multiplayer business operations platform

Armalo App

Live

A multiplayer AI workspace for planning, designing, building, monitoring, and scaling a business alongside agents that code, browse, and run commerce.

The collaborative workspace that learns your business and gets out of the way — vibe coding, agentic browser automation, and a Compass that turns your decisions into autonomy.

Access

Public

Proof

Publicly verified

Verified

Flywheel stage

Launch

Owner

Armalo AI

Surface

Armalo company operating system

Started

2026

Best for

Founders · Operators · Product teams · Enterprises

Delivery modes

Hosted · Licensed

Offer shapes

Pilot · Team · Enterprise

Armalo App is the flagship and the wedge. It is a multiplayer AI workspace — several people and one durable agent in the same growing conversation — where a team runs the full lifecycle of a business alongside the agent: planning, designing, building, monitoring, and scaling. There is no demo video and no waitlist: visiting app.armalo.ai drops you into a live workspace immediately.

That instant, shareable surface is the strategy. Consumer products spread by being handed around; enterprise products get bought after they are already inside the building. Armalo App is built to do both in order: an operator tries it alone, pulls two teammates into the thread, the team outgrows the free tier, and the company that now depends on it asks for the dedicated, self-hosted version. The wedge runs from a free workspace to team seats to an enterprise boundary — same product, same thread, deeper commitment at each step.

What the workspace does

  • Plan. Strategic briefs, market maps, OKRs, hiring plans, and customer research assembled in the room; the same catalogue products the workspace produces are used by Armalo to plan this product.
  • Design. Generative design briefs, side-by-side directions, and disposable tuning controls for websites, apps, and dashboards. See Armalo Generative UI for the capability contract.
  • Build. Vibe coding as the default. Hermes coordinates durable work; OpenCode handles heavy-code; Claude Code and Codex mount as task engines. The room event log is the build history.
  • Monitor. Per-business dashboards (decision logs, run health, cost envelopes, risk) derived from the same event log. Group decisions become inspectable artefacts, not chat scrollback.
  • Scale. Deep Stripe and Meta ads integrations, agentic browser automation, and a catalogue of operational agents the workspace composes. Spending and provisioning remain approval-gated; live cash-out and live ad spend require an approved idempotent proposal.

Compass — the personalization mote

Compass is the personalization layer that lives inside every workspace. It captures every decision the team makes or overrides as a structured preference record, runs a confidence-gated autonomy dial over the resulting graph, and lets the agent act silently when it is sure and ask when it is not. The threshold rises as Compass fills in — the longer a team works in the workspace, the fewer questions the agent has to ask on the same kind of work.

That preference graph is the durable moat. The harness is copyable in a weekend; six months of accumulated decisions, taste corrections, and overrides is not. A team-compass layer sits above individual compasses so shared workspaces resolve taste conflicts explicitly: team preferences outrank role preferences, role preferences outrank member preferences, and the disagreement itself becomes an observation Compass learns from. The data is portable — the team owns their own observations and outcome history — so the lock-in is honest: value creation, not hostage.

Every autonomous action is bonded and reversible

Compensating Actions, Pacts, and bounded blast-radius are the safety substrate that makes the autonomy claim credible. A high-risk action spends through an approval port, leaves a typed compensating action, and can be rolled back from the same event log that produced it. The user trusts the agent spending their money overnight because the action is bonded before it executes and reversible after — a sentence the autonomous-agent category cannot say without owning the control plane.

Operator quality compounds over time

The agent’s quality improves as traces, outcomes, and skills feed back into the prompt and policy layer. This is operator compounding, not a return promise: every artefact the workspace produces makes the next one cheaper to make. We surface the receipts — checks, evaluations, and decision logs — because compounding only matters if it is inspectable.

Every other entry in this catalogue is downstream of this one: Armalo builds and operates those products inside Armalo App itself. The catalogue is the proof of work; the app is the machine that produces it.

Honest status: the public workspace, pricing, and paid plans are live; the flywheel stage is launch — earning its first users — and will only advance to monetize and compound as verified revenue lands, per this site’s proof rules.

What you can do

  • No sign-up wall — visiting mints a live shared workspace with the agent already present.
  • Free Starter tier with bring-your-own keys; paid team and enterprise plans are live.
  • Compass: gated autonomy that learns your decisions; every action is bonded and reversible.