Your own AI Chief of Staff.
Built once. Runs forever.
Atlas is the always-on assistant we built for ourselves — it reads our email, calendar, calls and messages, briefs us every morning, and acts on our behalf. This is the complete blueprint to build your own. Every integration. Every step. Real code.
What you're building
The assistant a $200k Chief of Staff would be.
Atlas lives on a private server you own. It plugs into the tools you already use — Gmail, Calendar, Slack, Discord, Notion, your phone — and quietly turns the firehose into a searchable brain.
Every morning it tells you what actually needs your attention. During the day it answers from your own data, drafts the email, books the meeting, files the note. You approve with one click. Nothing sends without you.
This isn't a SaaS subscription. It's yours — your server, your keys, your data. Atlas is the blueprint, end to end.
The three jobs it runs on
One system. Three jobs. Every line of the guide serves one of them.
Capture
Pulls every connected app on a schedule into one knowledge base. Nothing slips through the cracks.
Surface
Reads it all, extracts what actually needs you, nudges you with one-click done or snooze.
Assist
Answers from your brain, manages tasks, drafts replies, books meetings — from one chat window.
A day with Atlas
What a real Chief of Staff would do. Quietly. All day.
Reads every email
Triages inbox, surfaces what needs a reply, drafts the response for one-click approval.
Runs your calendar
Books meetings, blocks deep work, reschedules conflicts before they hit you.
Listens to your calls
Transcribes, extracts action items, files follow-ups. You walk out of the call with the work already started.
Watches Slack & Discord
Pulls signal out of channel noise. Pings you only when something actually needs you.
Knows your Notion
Answers from your SOPs, offers, contracts — not from a generic model's hallucination.
Briefs you every morning
One message. Today's priorities, what slipped, what you owe people. Coffee-ready.
The stack
Every app, every key, every line of glue. No guessing.
You don't need to figure out which AI provider, which database, or which automation framework. We picked them, tested them, and wrote the integration code. You just follow the steps.
- DigitalOceanThe always-on server the whole system runs on.
- Anthropic (Claude)The accuracy + fallback model.
- GroqThe fast, cheap primary model.
- Slack APIWhere you talk to it. Where it nudges you.
- DiscordReads team chatter so you don't have to.
- Google CloudGmail + Calendar API access.
- Notion APIPulls your docs into the brain.
- Codex + Claude CodeHow you actually assemble it, fast.
What's inside
14 chapters. Every step laid out the same way.
First what the piece does. Then why it works that way. Then real code. Then what just happened. Then what to do next. Follow it top to bottom and you have a working assistant.
- 01 · The Concept — the three jobs and how the system fits together
- 02 · Foundations — security, secrets, server hardening, the rules that make it safe to run
- 03 · Stand up the server — DigitalOcean, accounts, keys, the day-zero setup
- 04 · Capture — pulling Gmail, Calendar, Slack, Discord, Notion on a schedule
- 05 · The vault — the markdown brain that becomes Atlas's memory
- 06 · Surface — the morning brief, nudges, and one-click done/snooze
- 07 · Assist — answering from your data without making things up
- 08 · Actions — draft, approve, send: outbound that's safe by design
- 09 · Calls & meetings — transcripts, action items, follow-ups
- 10 · Tasks & calendar — letting Atlas actually run your week
- 11 · The chat layer — how you talk to it (and what not to put in that channel)
- 12 · Backups & disaster recovery — never lose the brain
- 13 · Cost & scaling — what it actually costs to run, and how to keep it cheap
- 14 · Going further — the upgrades we'd build next
Built safe
The four rules that make an AI assistant safe to run on your real business.
Your keys, your server
Secrets in .env, server locked down with SSH keys + firewall. You own everything.
Approval-gated actions
Nothing sends on its own. Atlas drafts; you approve with one click. The rule that makes it safe.
No hallucinations
Atlas answers from your knowledge base or says it doesn't know. Never invents facts.
Backed up by design
The brain is markdown, mirrored to a private git repo on a schedule. One failure isn't fatal.
Who it's for
Atlas is for builders. Be straight with yourself.
For you if
- Founders and operators drowning in inbox, Slack, and calendar
- Anyone comfortable in a terminal — or willing to let Codex / Claude Code do it
- People who want to own their AI stack, not rent another SaaS
Not for you if
- — You want a one-click product with no setup
- — You won't keep your own keys or run your own server
- — You expect AI to do the thinking — Atlas amplifies you, it doesn't replace you
Atlas · Build Guide
$49one-time
No subscription. No upsell. The blueprint we use ourselves, written down so you can ship yours this weekend.
- The full 14-chapter Atlas Build Guide (PDF)
- Every integration step: Gmail, Calendar, Slack, Discord, Notion
- Real code you can paste into Codex or Claude Code
- Security, backup, and cost-control playbooks
- Free lifetime updates as we extend Atlas
Want us to build it for you instead? Book a build call.
FAQ
Common questions before you buy.
Do I need to be a developer?+
If you can edit a config file and run a terminal command, you can build Atlas. The guide is written to be assembled with Codex or Claude Code doing the heavy lifting — they handle the syntax, you handle the decisions.
What will it cost to actually run?+
Roughly the cost of a small DigitalOcean droplet plus your AI usage on Groq and Anthropic. We use cheap/fast Groq as the primary model and fall back to Claude only when accuracy matters. Most users keep it under $30–60/month all-in.
Is my data going anywhere?+
No. Atlas runs on your server. Your knowledge base lives in markdown files you own, backed up to your own private git repo. The only thing leaving the box is the prompt you send to the model.
Can Atlas send emails without me?+
Only if you tell it to. By default every outbound action is approval-gated — Atlas drafts, you approve with one click. That single rule is what makes the whole thing safe to run.
What if I get stuck?+
Every chapter ends with what just happened and what to do next, so you always know where you are. If something breaks, the guide tells you exactly what to ask Codex or Claude to fix.
Do I get updates?+
Yes — lifetime updates included. When we extend Atlas (new integrations, new agents) you get the new chapters.
Instant PDF · Lifetime updates · No subscription