The AI operating system for ecommerce, running on your own machine.
A crew of agents that watch your store, explain what is wrong, and fix it. Every action staged, human-approved and reversible. Self-host in minutes: one command, and it runs in your browser. Source-available; your data never leaves.
Already building in n8n? Import your workflow and compile it into an agent in one click.
One command · runs at localhost · SQLite on disk · no telemetry · bring your own LLM key
Scout, Kai, Rai & Memo · a live audit loop · detect → fix → remember
Your n8n workflows, compiled into agents.
Paste any n8n workflow JSON and Vortex Runtime compiles it into a working agent: triggers become schedules, HTTP and app nodes become real tools, code nodes run verbatim in a sandbox, and AI prompts become the agent's objective. Import your n8n credentials into the encrypted local vault, picked up automatically. Or skip n8n entirely and build an agent from scratch.
Import, don't rebuild
Workflow JSON in, agent out. Nothing to re-wire, nothing to re-learn.
20 community templates
Start from a pre-converted n8n template. The simple ones run as-is; the richer ones are a head start.
No n8n required
Build an agent in a six-step wizard, or let the built-in crew audit and fix your store out of the box.
Not a SaaS login. A local server you own.
Launch it with one command and Vortex Runtime starts a small web server on your own machine, then opens the full crew office in your browser at 127.0.0.1:8420. Everything lives in local SQLite. Deploy it headless with Docker for on-prem and it works the same.
One-command launcher
Opens the office at a tokenised localhost URL. The fastest way to see your crew work.
From a checkout
Clone, install, run. Read every line. The audience for fair-code is technical, and welcome to be.
Headless / on-prem
The same artifact self-hosts on a server behind your firewall for enterprise.
Detect · Explain · Fix · Remember.
The office is laid out as the pipeline, not a feature grid. Four named agents do the work out of the box. You can watch every one of them, and rename them if you like.
Audits your store and every connected source. Finds what is broken, risky, or quietly costing money. Deterministic, no model needed.
Runs the auto-approved, provably-safe fixes: schema, robots, llms.txt, meta. The fixes that cannot go wrong.
Prepares the reversible changes that need a human go. Staged, diffed against a real page, waiting for your approval.
Files what every run teaches about your store. The memory that makes month twelve smarter than month one.
Nothing touches your live store without you.
Every change is staged, never live until you approve it, and exactly reversible. You see what each agent will do, why, and how to undo it, with signed provenance on every action. The free runtime teaches the same trust model the paid platform runs on.
Preview before live
Changes run against a working copy and show a real before/after diff. Nothing writes to production unannounced.
You hold the go button
Ungated fixes auto-apply; anything with risk waits for a human. Write access never means autonomy.
One-click rollback
A restore point is verified before the first write. Proven byte-exact on live stores.
Choose how much stays inside your walls.
Same crew, same skills, same governance: you decide where the reasoning runs. From cloud-assisted with your own key, all the way to no model ever called.
Cloud, bring-your-own-key
most teamsLocal, deterministic skills; escalate the hard reasoning to a model on your own API key. Cheapest to run.
On-prem, local AI
no data leavesEverything inside your walls. A local model only. No third-party model is ever called.
On-prem, your own model
frontier, insideEscalate to a large model you host yourself. Full capability, zero exposure.
Zero-LLM
strictestDeterministic skills only: audits, schema, feeds. No model runs at all.
One honest exception, shown before you choose: Google Ads calls route through a thin Vortex IQ proxy (a Google developer-token rule, not ours).
Fair-code. Not open source, and we will always say so.
Source-available means you can read it, run it, and change it. Fair-code means one honest limit: you cannot take it and sell it back to the market as a hosted service. It is the model n8n, Sentry and others run real businesses on.
We call it "fair-code" or "source-available", never "open source". Claiming pure OSS when it is not burns the trust this is built on.
Start free on your machine. Graduate when you want it hosted.
The runtime is genuinely useful on its own, never crippled. You move up when you want it always-on, unifying your whole stack, with your team. A graduation, not a paywall.
Vortex Runtime
fair-code · self-hosted · free- The full agent framework, connectors and the crew office
- Deterministic audits, scans and the staged fix loop
- n8n import and the agent studio
- One store, run locally; bring your own LLM key
- Runs offline; your data stays on your machine
Vortex Cloud
hosted · always-on · from the AIOS ladder- Hosted with no infra; Vortex-run models for heavy agents
- Always-on scheduling and monitoring
- Cross-platform data unification and memory at scale
- Teams, governance and multi-store
- Dedicated support
Be first on Vortex Runtime.
We are inviting a small group of merchants and builders to the private beta ahead of the public release. Tell us where to reach you.
Vortex Runtime is built and in private validation. We will notify you the moment the public, source-available release is ready. No spam, no "launched" before it is true.
