Agent Inbox
Give your agent its own address.
Aghata doesn't ask for the keys to your mailbox. It gets a real address of its own — one that can work an outcome over email: send, wait for the reply, follow up, and loop until the goal is met, with you on the approvals.
The address
Not access to your inbox. An inbox of its own.
Every other email assistant wants OAuth into your mailbox. Aghata takes the other road: its own address, on its own domains, with your oversight built into every send.
A real address, provisioned for you
Vendors, candidates, and customers write to the agent directly. Every conversation lives in a normal inbox you can open, search, and jump into whenever you want.
Two domains, two reputations
Conversation and marketing send from separately registered domains, each with its own standing. A newsletter can never taint the address your vendors reply to.
Deliverability handled
Sending authentication is configured and managed for you — the plumbing that decides whether email lands is not your problem to babysit.
Email goals
It sends. It waits. It follows up. It knows when it's done.
An email goal is a durable session working an outcome over correspondence. The clock is the other side's reply — and waiting is free.
The goal plans its opening move and sends — or parks a draft for your sign-off first, depending on the autonomy you chose.
The session pauses for input and goes fully dormant. No polling, no idle agent, no cost — for an hour or a week.
The reply lands, is matched to its thread, and wakes the exact session that was waiting — with all its context intact.
The goal evaluates: met, keep working, or needs you. It follows up, escalates to you, or closes out with the thread as the record.
"Chase the invoice until it's paid" is a session that might span two weeks and cost you nothing while it waits — because between sends, there is nothing running.
Your hand on it
Autonomy is a knob, not a leap of faith.
You decide how much rope every goal gets — and you can take a thread back, or hand one over, at any moment.
Manual, draft, or auto. Per goal: it can propose and wait, draft with Send / Edit / Discard in the thread, or reply autonomously to correspondents you've allowed.
Allowed senders. Autonomous replies can be restricted to a list you control. Everyone else's messages park the goal for your review.
Hand to Aghata. Any thread you started can become a goal mid-conversation. It reads the history and drafts the next reply, in-thread.
A banner that tells the truth. Every AI-worked thread shows its state — working, waiting for reply, needs you, done — with Reply myself and Stop always available.
Typed pauses, same as everywhere. A goal that needs a decision parks for input; one about to do something consequential parks for approval. Inbox work follows the same contract as every session.
The thread is the audit. Everything sent and received is right there in the conversation — no hidden actions, no side channel.
Campaigns
Marketing with a human on the send button.
The same inbox scales to audiences — with the guardrails that make volume email survivable: isolation, suppression, and consent as records.
AI prepares. A person enqueues.
A campaign drafted by a session stays a draft — linked back to the run that wrote it — until a human presses send. There is no autonomous path to a thousand inboxes.
One suppression list, checked every send
Bounces, complaints, and unsubscribes feed a single authoritative list consulted on every outgoing message — conversational or campaign. A global entry blocks both engines.
Consent is a record
Each contact carries its lawful basis, source, and timestamp, and every consent change is logged. Delivery, opens, clicks, and unsubscribes report back per campaign.
Questions
The parts people ask about first.
Does Aghata read my Gmail?+
No — and that's the design, not a limitation. Aghata gets its own real email address on its own domain. People write to it, it writes back, and you supervise. Your personal mailbox never enters the system.
What's an email goal?+
An outcome you hand to the inbox: 'get three quotes for the office move,' 'chase this invoice until it's paid.' Aghata sends, waits for replies, evaluates whether the goal is met, and loops — as one durable session you can open, steer, or stop at any point.
What does waiting cost?+
Nothing. A goal that's waiting on a reply is a parked session — no agent loops in the background. The incoming email itself wakes the run, whether that's in an hour or a week.
Will it ever send something I haven't seen?+
Only if you tell it to. Autonomy is one knob — manual, draft, or auto — and in draft mode every outgoing message parks as a draft with Send, Edit, and Discard in the thread. You can also restrict autonomous replies to senders you list.
Can it take over a conversation I started?+
Yes. 'Hand to Aghata' converts any existing thread into a goal: it reads the transcript, drafts the next reply in-thread, and works the conversation to done — with a banner showing whether it's working, waiting, or needs you.
How do marketing sends stay safe?+
Campaigns run on a separate domain from conversation, so reputations never mix. AI-prepared campaigns are drafts until a human enqueues them, one authoritative suppression list is checked on every single send, and consent is stored as a record — basis, source, timestamp — not a checkbox.
Chapters of one system
Give it the outcome. Keep the control.
Start free in the product, or talk to us about how your team would use it.