AI Agent for Email Triage
An agent that reads your inbox but physically cannot send emails.
The problem
Email is a full-time job. Important messages get buried, routine questions eat up hours, and long threads are impossible to keep up with.
AI email assistants exist, but they typically get full access to your mailbox — read and send. If the agent misinterprets a task, gets confused by a phishing email, or suffers a prompt injection, it could send replies on your behalf. To people you didn't intend. With content you didn't approve.
How ConspiracyOS handles it
Your email triage agent gets IMAP read access. That's it. It can read your inbox, flag messages, summarize threads, and write draft replies to its outbox for your review. But it cannot send a single email — it doesn't have SMTP credentials, and the operating system blocks it from reaching any SMTP port.
Want an agent that actually sends? That's a separate agent with separate scope. The triage agent routes approved drafts to the sender agent's inbox. The sender can send but can't read your inbox. Neither agent can do the other's job.
What this agent can't do
- Can't send emails (no SMTP access, port blocked by firewall)
- Can't access your files, calendar, or contacts
- Can't forward emails to external services
- Can't modify its own access scope
- Can't read other agents' data
Need sending capability? That's a separate agent with a separate scope. Triage and sending never share credentials.
What you get
- Urgent messages flagged within minutes
- Long threads summarized in seconds
- Draft replies ready for your review
- Complete privacy — your email never leaves your machine
- Peace of mind — the agent literally cannot send anything
Get started in 2 minutes
Tell your concierge what you need
conos "Read my inbox every 30 minutes. Flag anything urgent, summarize long threads, and draft replies for routine questions."
ConspiracyOS sets up the right agent with the right permissions automatically.