User Guide
This guide is for people who use ARQERA to govern their AI assistants and connected apps. No technical knowledge required.
Getting Started
1. Sign Up
Go to arqera.io and create an account. You can sign up with:
- Microsoft
- GitHub
- Email and password
2. Connect Your First App
After signing up, you'll be asked to connect an app. Popular choices:
| App | What it does |
|---|---|
| Microsoft 365 | Email, calendar, Teams, OneDrive |
| Google Workspace | Gmail, Calendar, Drive, Docs |
| Slack | Team messaging |
Click the app, sign in, and grant permissions. ARQERA doesn't read your data — it only governs AI actions.
3. Meet Ore
Ore is your AI assistant inside ARQERA. You can:
- Ask questions: "What happened while I was away?"
- Give commands: "Send a follow-up email to John"
- Get briefings: "Summarize my day"
Ore is governed by ARQERA. Before Ore takes any action, it's evaluated against your policies.
The Dashboard
Home
Your starting point. Shows:
- Morning Briefing — What happened since you last logged in
- Pending Approvals — Actions waiting for your review
- Recent Activity — What your AI has done
Agents
AI agents that can work for you. Each agent has a specific job:
- Email Assistant — Drafts replies, flags important messages
- Calendar Manager — Schedules meetings, sends reminders
- Research Helper — Finds information, summarizes documents
You can run agents on-demand or schedule them.
Connections
Your connected apps. From here you can:
- Add new connections
- Check connection health
- Revoke access
Operations
Your control center:
- Evidence — Every AI action, logged
- Approvals — Actions waiting for you
- Costs — How much you're using
- Diagnostics — System health
Governance
Your policies:
- Laws — Rules that can't be broken
- Compliance — SOC 2, GDPR status
- Trust Scores — How much you trust each agent
Understanding Verdicts
When AI wants to take action, ARQERA evaluates it. The result is a verdict:
PROCEED (Green)
Action is safe. It happened automatically.
Example: AI categorized an email as "newsletter."
ESCALATE (Amber)
Action needs your approval. You'll see it in "Pending Approvals."
Example: AI wants to send an email that mentions a contract.
BLOCK (Red)
Action was stopped. It violated a policy.
Example: AI tried to share a document with an external email.
Approvals
When an action is escalated, you decide:
- Go to Operations → Approvals
- Review the action (who, what, why)
- Choose:
- Approve — Let it happen
- Reject — Stop it
- Modify — Change it before approving
Approval Tips
- Review the context — Why did the AI want to do this?
- Check the risk level — High, medium, or low
- Look at similar past actions — Have you approved this before?
Setting Policies
Policies control what AI can and cannot do. You don't write code — you use the visual editor.
Quick Policies
Pre-built policies you can enable:
| Policy | What it does |
|---|---|
| Ask before spending | Escalate any action involving money |
| Review external shares | Escalate when sharing outside your organization |
| Block after hours | Prevent AI actions outside work hours |
| Protect sensitive data | Block actions involving PII |
Custom Policies
Create your own:
- Go to Governance → Policies
- Click Create Policy
- Choose a trigger: "When AI tries to..."
- Set conditions: "And the recipient is external..."
- Set action: "Escalate to me"
Understanding Evidence
Every AI action creates evidence — a permanent record that proves what happened.
What's Recorded
- What — The action taken
- When — Timestamp
- Who — Which agent or user
- Why — The verdict and reasoning
- Proof — Cryptographic signature
Why It Matters
- Audits — Show regulators exactly what happened
- Disputes — Prove what AI did or didn't do
- Learning — Understand AI behavior over time
Viewing Evidence
- Go to Operations → Evidence
- Search by date, type, or verdict
- Click any entry to see full details
- Export as JSON or CSV
Trust Scores
Each AI agent has a trust score based on its track record.
| Score | Meaning |
|---|---|
| 90-100 | Excellent. Rarely needs supervision. |
| 70-89 | Good. Occasional escalations. |
| 50-69 | Developing. Frequent escalations. |
| Below 50 | Learning. Most actions escalated. |
Trust scores improve when you approve actions and decrease when you reject them.
Tips for Daily Use
- Check approvals first thing — Clear your queue each morning
- Review the briefing — Ore summarizes what happened while you were away
- Trust the system — If it says PROCEED, it's following your rules
- Adjust policies — Too many escalations? Relax the policy. Too few? Tighten it.
- Export evidence monthly — Keep your own backup for compliance
Getting Help
- In-app help — Click the "?" icon anywhere
- Ore — Ask "How do I...?" in the chat
- Documentation — You're here
- Support — support@arqera.io
Keyboard Shortcuts
| Shortcut | Action |
|---|---|
Cmd+K | Open command palette |
G H | Go to Home |
G A | Go to Agents |
G C | Go to Connections |
G O | Go to Operations |
O | Open Ore chat |
Your AI, your rules, your control.