Install

Ways to plug into Aurais

Start with whichever surface is closest to how your agents already run. Evidence always classifies server-side; the catalog never ships to a client.

Command-line scanner

shipping

One command. Scans your device for AI agents, streams findings back to aurais.net in real time. Three run paths — pick whichever matches how your machine is set up.

1. Paste one line (zero install, fastest path)

macOS / Linux / WSL — bash:

curl -sL https://aurais.net/scan.sh | bash

Windows — PowerShell:

iwr https://aurais.net/scan.ps1 | iex

No Node, no npm, no download. The script mints a pair session, prints a verification code that matches what your browser shows, then collects quick evidence using shell builtins only (package manifests, .env, MCP configs, running AI processes, listening ports, and on Windows Get-AppxPackage for Store apps). All classification happens server-side.

2. Download signed binary (deep scan, no runtime required)

Windows (x64)

55.0 MB

macOS (x64 / Intel)

62.3 MB

Linux (x64)

71.0 MB

v0.1.0. Binaries bundle Node — no separate runtime required. Code-signing pending: Windows SmartScreen may warn; macOS will ask you to approve on first run. Verify the SHA-256 above matches what you downloaded if you're careful.

2. Have Node 20+? Run the script directly

node c:/voriongit/vorion-find/dist/index.js --pair=NEW --depth=standard

Uses the CLI source in-tree. Zero runtime deps; uses native fetch. Evidence trimmed to 8 KB per item client-side; classifier runs server-side.

3. Via npm (coming soon)

npx vorion-find --pair=NEW --depth=standard

Publishes once the voriongit/vorion-find GitHub remote is created and a release is tagged.

Flags on all paths: --pair=NEW to mint a session from the CLI, or paste the sessionId:pairToken from the browser's Scan this device panel for mutual-code verification. --depth accepts quick | standard | deep | fulltilt.

VS Code extension

preview

Workspace scanner, trust filesystem mount, agent dashboard, proof-chain diff. 'Send to aurais' ships in Stage M.

vorion://agents/{id}/ filesystem mount exposes identity, governance policies, behavior, attestations, and integrity per agent.

Quick / Standard / Deep workspace scan via the command palette (Cmd-Shift-P → Vorion: Discover Agents).

Live dashboard panel with gauges + sparklines + integrity dials for every tracked agent.

Current extension lives in-tree at c:/BAI/vorion-vscode-extension. Marketplace publication pending. Stage M adds a one-click "Send findings to aurais.net" action that pairs with the same SSE channel the CLI uses.

GitHub App

soon

On PR open: scans the repo for ungoverned AI, posts a check with findings, offers one-click onboard.

Stage M deliverable. Repo-level scan via the GitHub Checks API; status check reports governed/ungoverned counts; optional auto-comment with the scare-UI summary and a one-click onboard button. Uses the same proprietary fingerprint catalog as the web + CLI paths via /api/detect.

Browser extension (ChatGPT / Claude.ai observer)

soon

Observe-tier attestation for conversations inside chatgpt.com and claude.ai. Closed-source agents cap at Observe + Mitigate — no Protect.

Stage M deliverable, shipping only for browsers we can actually honor the trust contract on. ChatGPT consumer and Claude Projects do not route LLM calls through our process — the honest ceiling is Observe tier (see plan §3a and the homepage copy). No Protect claims for closed-source agents.

Everything lands in the same mission

Findings from any surface ride the same SSE channel per-session and land in the same Mission Control watchlist. Nothing forks by platform.

Run a scan nowGenerate a snippet