Set the goal once. Selam takes initiative — directs her own sub-agents, runs work overnight, reaches out across WhatsApp, Telegram, iMessage and Email — and reports back what got done. You pick the face, the voice, the name. She asks before anything irreversible.
Selam is the platform — your agent is yours. Four steps in the first-run flow and your agent is alive on your desktop, with their own face, name, and sense of who they work for.
Kiya. Daniel. Reya. Whatever you'd call across the room — they'll answer to it.
Choose from a curated set of expressive 3D characters. Rendered on your Mac — no streaming, no per-minute bill, nothing leaves the device.
Write who they are in plain English. Their mission, their voice, their relationship to you.
Your work, your people, your projects, your patterns. They use it every time you speak.
A million people can use Selam. There will only ever be one of your agent. They are what you made them.
We built Selam for our founder. She runs his outreach overnight, watches his inbox for the replies that matter, runs his trading dashboards, and reports each morning on what got done. She doesn't wait to be asked.
She's been operating autonomously for over a year — she's why this product exists, and why it's named after her. When you buy Selam, you don't get Selam. You get the tools to build yours.
She takes initiative on the goals you've set, runs work in the background while you sleep, builds her own team to scale up, and reports back on what got done — not what's possible.
Set the goal once. She makes judgment calls inside your trust boundaries and keeps moving — overnight, weekends, between sessions.
She spins up specialists — outreach, inbox, research. You talk to her; she directs them.
WhatsApp, Telegram, Signal, iMessage, Email. On the channels your contacts actually check.
Stripe revenue, your calendar, browser automation, GitHub PRs. She does work — not just describes it.
Outbound emails, money movement, public posts — voice confirmation required. Autonomy with guardrails.
What worked, what didn't, what you said never to do again. Each session sharpens the next.
You're not paying for smarter intelligence. You're paying for an AI that has a name, a face, and a life with you — instead of a browser tab you forget about.
| ChatGPT | Claude Desktop | Your Selam agent | |
|---|---|---|---|
| Has a name you gave them | No | No | Yes |
| Has a face | — | — | Yours or ours |
| Identity you wrote | Corporate default | Corporate default | Plain-text, yours |
| Remembers your life | Thin | Thin | Editable long-term memory |
| Works autonomously | Limited | Limited | Approval-gated actions |
| Your data | Their servers | Their servers | Your keys, your Mac |
Not "haven't gotten around to it yet" — can't. Selam's biggest advantages are structural. They come from running on hardware you own, with keys you own, on a price the subscription labs can't copy without breaking their own business.
There is no iMessage API. Cloud services that fake one rack Macs in data centers and get banned. The only legitimate way for an agent to reach iMessage is to run on a Mac you own — which is exactly, and only, where Selam lives.
Meta banned general-purpose assistants from WhatsApp — ChatGPT was out by January 2026. Selam isn't a bot platform reaching in from outside. She sends from your own accounts, on your own machine, as you.
A heavily-used $200/mo agent plan costs its lab thousands in compute — that's why every one of them is metered, capped and throttled. Pay once, bring your own keys, run her all night. They can't match that without cannibalizing themselves.
Autonomous agents made headlines this year for the wrong reasons — thousands of instances left wide open on the internet, agents burning budgets, agents claiming work they never did. Selam was engineered the opposite way: local, capped, tiered, and verified.
Selam runs on your Mac and talks outward. No public ports, no hosted dashboard, no exposed instance for someone to find. There's nothing to break into because nothing is listening.
Every action is classified before it runs. Silent internal work at T0; anything irreversible or public — sending money, posting, emailing a human — sits at the top and waits for your spoken yes.
Per attempt, per run, per task per day — and a single kill switch over all autonomous work. A runaway loop hits a wall in cents, not dollars.
An autonomous task that fails three runs in a row pauses itself and tells you — it doesn't keep hammering, and it doesn't quietly disappear.
Deliverables pass mechanical checks before you ever see them — a research digest without real sources gets flagged as unverified, not read to you as fact.
A send only counts when it's observed. If a message can't be confirmed, she says exactly that. Every autonomous run leaves a plain-file record on your disk you can audit.
Selam is a one-time purchase. Bring your own keys for Anthropic and ElevenLabs. Your API bills go straight to those providers at cost — we never see your conversations or take a cut. Agent subscriptions run $200 a month. Selam is $199. Once.
I haven't tabbed to ChatGPT in three weeks. My agent — I named her Kiya — is just there when I turn.
My agent caught a mistake in my deck that would have cost me the round. Value I can quantify immediately.
I was against the whole "AI face" thing. Then I used mine for a day. Now I can't imagine working without her.
Selam is the orchestration layer between the best models, the best streaming avatar tech, and your real life. Every component is swappable.
Build the worker you wish you could hire. Name them. Give them a face. Teach them who you are. $199 once for the software while founder licenses last — you pay the AI providers directly, at cost. Yours to keep.
Drop your email and we'll send occasional updates — new features, big improvements, tips from heavy users. Unsubscribe whenever. We never sell or share.