Most AI tools start every conversation cold. You explain who you are, what you sell, who your customers are, what tone to use, what not to say. Five sessions later, you're still pasting the same brief into a new chat window. It's the IKEA furniture of AI — convenient, generic, lifeless.

Ergora Remote is the answer to that. It's a small desktop agent that runs on your work machine — your home laptop, your office desktop, your studio Mac. With your explicit per-app permission, it observes how you actually work: the emails you write, the docs you draft, the Slack threads you start. Not to surveil you. Not to send your data anywhere. To learn how you communicate, so the AI working for you in the cloud sounds like you.

The "second brain" that actually knows you

The phrase "second brain" gets thrown around a lot. Most of the time it means a notes app that organises your scribbles. We mean something different.

Ergora's Seat Brain is the layer of memory that's strictly yours on a given project. It captures your voice fingerprint — the median sentence length you write, the openers you favour ("Hi name," vs "Hey name —" vs "name,"), how often you use contractions, your hedge words, your sign-off. It captures your decision style — do you want one recommendation or three options with tradeoffs? It captures your working patterns — when you focus, when you escalate, what you ignore.

In its chat-only form (today), the Seat Brain is calibrated from your conversations with the Ergora agent. That's fine. It works. But it's a slow drip — you have to chat for weeks before it has enough samples to be confident.

With Ergora Remote, that calibration period collapses to days. The agent on your machine processes the artefacts you create as part of your normal work — locally, on your device, never uploading the raw text — and syncs only the distilled features (sentence length distributions, opener counts, formality scores) to your Seat Brain in the cloud.

By day 14, the agent drafting your sales follow-ups doesn't sound like you. It sounds as you.

Three places, one continuous brain

Most people have at least two work environments. Home machine, work machine. Founder with a laptop and a desktop. Designer with a Mac for client work and a Windows box for gaming-during-lunch-breaks.

Ergora Remote stitches them together. Each device gets its own role — home, work, laptop, studio, server — and your Ergora chat can route requests to the right one:

  • /home find spec.docx — searches the home machine for the file
  • /work check my email — opens your work inbox
  • /laptop run npm test — runs the test on the laptop
  • /find quarterly budget — searches every connected device

The agent in the cloud chooses which machine to talk to based on context. Files don't get copied to the cloud. Your secrets don't move. The agent just routes to the right place + reports back.

But more than the routing, it's the knowledge that compounds. Your Seat Brain absorbs how you write at home (often more casual, exploratory) versus how you write at work (often more measured, decisive). The agent learns to draft an out-of-hours email in your home tone, a board update in your work tone, a client proposal in your client-work tone. Same brain, calibrated to context.

What it's deliberately not

Three things we made early decisions on:

1. Raw artefacts never leave the device. When the local agent observes an email you sent, it extracts the structural features — sentence count, length distribution, opener pattern, sign-off — and syncs only those. Your actual email content stays in your local Mail.app database. We can't see it. Your team can't see it. Even you can't pull it back out of Ergora — it was never there to begin with.

2. Per-app consent, with a pause button. You don't enable "Remote" wholesale. You enable specific apps: Mail, Slack, Docs, Browser. Each one is a separate consent. Each one has a pause toggle. Each one has a kill switch. We've biased every default toward the conservative choice.

3. No correlations across users. Your Seat Brain is yours alone. Your teammates have their own. Your business's wiki is shared, sure — but voice features stay personal. The agent never reveals "based on how Tom writes" to a colleague. It just produces drafts in their own voice when they're the one asking.

What this means for your work, in three scenarios

Scenario one — Sunday evening, your kitchen table. You think of a cold-outreach idea. You open Ergora Chat on your laptop, type "draft a cold email to Tom at Acme — he just raised a Series B, fits our ICP." Without Remote, the agent drafts something competent but generic. With Remote, the agent has weeks of evidence about how you cold-outreach: the hook structures that have been replied to, your usual specificity level, the line length you keep emails at. The draft is one button-press from sendable.

Scenario two — Monday morning, your office. Daily briefing in your inbox at 7:45am. The agent worked overnight. Three things it noticed: (1) two of last week's draft emails were heavily edited before sending, and the diff suggests you've shifted to a tighter sign-off — would you like to update your voice profile? (2) Your hero product changed in your wiki on Friday but two ad campaigns still reference the old SKU — flag for review? (3) A competitor of yours just announced something relevant — newsjacker has a draft response ready.

Scenario three — Wednesday, on a flight. You're offline. You can't reach the cloud. But your local Remote agent has the last 30 days of your wiki, voice profile, and cached chat. You ask it to draft a follow-up to a meeting from yesterday. It works locally — same voice, same business context, same precedence rules — and queues the draft to sync the moment you reconnect.

The trust contract

We mean it when we say we built Remote so you stay in control:

  • You can see everything. A "Recent learnings" panel shows every voice-feature delta the local agent has captured + synced.
  • You can wipe it instantly. One click locally deletes everything Remote has stored, plus the corresponding cloud-side voice profile.
  • You can audit every grant. Per-app consent is logged with a timestamp; revoking is one click.
  • You can run it air-gapped. Premium tiers will support fully on-device mode where no features sync — the cloud agent uses only what you give it in chat.

If at any point you want out, you're out. Cleanly. Within seconds.

Why this matters

The premise of every general AI tool is the same: a brilliant generalist that you train, every session, to your specific situation. The premise of Ergora is the opposite: a system that already knows your situation, and gets sharper at understanding it the more you work.

Remote is the missing piece that takes "knowing your situation" from theoretical (you described your business once during onboarding) to practical (the agent has watched you work for 30 days and the drafts are indistinguishable from yours).

It's the difference between an intern who's been with you a week and one who's been with you a year. We're trying to build the year-long colleague — without you having to wait a year.

Coming with launch

Ergora Remote ships alongside the public launch of the Ergora platform later this month. It runs on macOS first (Apple Silicon, Intel), Windows next, Linux third. The base agent is included with every Pro tier and above. On-device mode is bundled with the Enterprise tier (target Q3 2026).

If you want to be on the early-access list, sign up at ergora.cloud — accounts created before launch get the standard 50% off the first month, and Remote access on the day it ships.