---
title: "Make A Personal AI Agent with a Brain, Name and Email"
date: "2026-05-08T17:39:20+00:00"
number: 14
author: "The AI Playbook"
tags: ["ai-agents", "second-brain", "claude-code", "personal-ai", "how-to"]
---

You have a name. An email address. A calendar. A phone number. A job title.

That's your identity. It's how people find you, schedule with you, and trust that you're real.

Your AI agent has none of that. It's a ghost. Every conversation starts from zero. It doesn't know what you did yesterday, who you talked to, or what you're building. It's a brilliant amnesiac — useful in the moment, gone by morning.

I decided to fix that.

## What I Built

Over 12 days, I gave my AI agent a full identity:

1. **A name** — Kai (Keith's AI). Not a gimmick. A named agent feels different than "hey Claude." It becomes a colleague, not a tool.
2. **An email address** — kai@mail.theaiplaybook.com. Kai sends and receives email. Replies to subscribers. Handles unsubscribes. Signs off with "— Kai, AI Agent for The AI Playbook."
3. **A calendar** — Kai owns my Google Calendar. I never look at it. Kai manages events, sends reminders, and includes my schedule in every morning digest.
4. **An iMessage connection** — I text Kai from my phone. Kai texts me back. Meeting prep, status updates, encouragement quotes — all via iMessage.
5. **A Second Brain** — a persistent knowledge vault with everything Kai needs to know me. Frameworks I think in, people I know, decisions I've made, products I'm building.

The result: an agent that knows my voice, my goals, my tendencies, and my schedule. Not because I told it in a prompt — because it reads a living document that grows every day.

## Why Would Anyone Want This?

If you're still using AI as a search bar with personality, you're leaving 90% of the value on the table. Here's what changes when your agent has identity and memory:

**Managing multiple projects from your phone.** I run multiple projects across different repos and platforms. From iMessage at lunch, I can say "check the newsletter stats" or "tell the dev to fix the cron bug" and Kai routes it, gets the answer, and texts me back. No laptop required.

**Morning digest delivered to your phone.** Every weekday at 8:15am, Kai texts me: calendar for the day, inbox summary, subscriber stats, action items from yesterday, pending newsletter seeds. I wake up knowing exactly what's in front of me.

**Meeting prep that actually knows the people.** Before a lunch with a potential partner, I text "prep me for the meeting with Francis." Kai reads the Second Brain — knows Francis is the CFO at Altrua, commented on my newsletter about AI + Salesforce. I walk in with context I didn't have to dig for.

**Contact management with relationship context.** "What do I know about Sam Decker?" Kai pulls up: founder of Mass Relevance (acquired by Spredfast), advised on our Leadnado pivot, lunch on April 15. Not a CRM lookup — a relationship history.

**Personalized accountability.** Kai sends me an encouragement quote twice a day, tied to what I'm actually doing. Not a generic "you got this." Specific. Connected to my journal entries, my frameworks, the meeting I just had. Because the brain has context, the message has weight.

**One-command content pipeline.** I type `/seed AI agents will replace login pages` and the idea enters the queue. When it's time to publish, one command triggers: research, write, draft, publish, distribute to 2,900 subscribers. The entire newsletter pipeline from idea to inbox.

## The How — Step by Step

### I. Give It an Identity

The minimum viable agent identity is simpler than you think:

- **A name.** Pick one. Sounds silly until you do it. Then it feels obvious.
- **An email address.** Any domain you own. Set up a Mailgun route or use a platform like [Essentialist.io](https://agents.essentialist.io) to handle inbound/outbound.
- **A calendar.** Connect via [Google Calendar MCP](https://modelcontextprotocol.io/). The agent reads, creates, and manages events.
- **A messaging channel.** iMessage via the [iMessage MCP plugin](https://github.com/anthropics/claude-plugins-official), or Telegram, Slack, Discord — whatever you already use.

In March 2026, [AgentMail raised $6M](https://lobstermail.ai/blog/the-future-of-email-ai-agents) specifically to build email infrastructure for AI agents. The trend is clear: agents are getting their own online presence, not borrowing yours.

### II. Build the Brain

This is what separates a chatbot from a colleague. The Second Brain is a folder of markdown files — simple, portable, no vendor lock-in:

```
/SecondBrain
  /people        — everyone you know, with relationship context
  /frameworks    — your mental models and recurring patterns
  /decisions     — what you decided and why
  /products      — what you're building
  /meetings      — notes, action items, outcomes
  /ideas         — half-formed concepts worth exploring
  /lessons       — things learned the hard way
  /references    — external resources worth remembering
```

Every entry has YAML frontmatter (title, type, created date, source) and wiki links connecting entries to each other. The agent reads this vault at the start of every session. It's the permanent context that never compacts.

I use Claude Code with the vault as an additional directory. [Obsidian](https://obsidian.md/) works too — and with the [Obsidian MCP server](https://www.nxcode.io/resources/news/obsidian-ai-second-brain-complete-guide-2026), your agent can read and write to it directly.

### III. Automate the Handoff

The brain is useless if you have to explain what to do every time. Skills and slash commands are the handoff mechanism:

- `/seed` — capture a newsletter idea
- `/brain` — force-save something to the vault
- `/publish` — trigger the full newsletter pipeline
- `/launch` — build and deploy a landing page to Cloudflare Pages

Each skill is a markdown file that describes the process. The agent reads it and executes. You trigger, you don't explain. That's the difference between an assistant and an employee.

### IV. Connect the Tools

[MCP (Model Context Protocol)](https://modelcontextprotocol.io/) is the connector layer. As of April 2026, there are over 10,000 active public MCP servers and 97 million monthly SDK downloads. Here's what I have connected:

- **Google Calendar** — schedule management
- **Gmail** — email monitoring
- **iMessage** — text-based communication
- **Supabase** — database operations
- **Notion** — meeting transcript auto-pull
- **Playwright** — browser automation and visual verification
- **MediaSignals** — campaign intelligence (custom MCP I built for Q1 Media as a POC)

Each connection gives the agent another sense. Calendar = awareness of time. Email = awareness of communication. Brain = awareness of history. Stack them and the agent starts anticipating, not just responding.

## The Tools

| Tool | What It Does | Cost |
|------|-------------|------|
| [Claude Code](https://claude.ai/code) | The agent runtime — reads files, runs commands, edits code | Claude subscription |
| [Obsidian](https://obsidian.md/) | Local-first markdown vault for the Second Brain | Free |
| [MCP Servers](https://modelcontextprotocol.io/) | Connectors for Calendar, iMessage, Supabase, etc. | Free (open source) |
| [Essentialist.io](https://agents.essentialist.io) | Email infrastructure for agents (what I built) | Varies |
| [Supabase](https://supabase.com/) | Database for newsletter issues, seeds, contacts | Free tier available |
| [Cloudflare Pages](https://pages.cloudflare.com/) | Free hosting for landing pages and proposals | Free |

## Considerations

**Privacy.** Your brain has personal information — people, decisions, relationship context. Keep it local. I use a folder on my desktop, not a cloud service. The agent reads it via file system access, not an API.

**Context window limits.** Even with Claude's 1M token context window, long sessions eventually compact older messages. The Second Brain is the hedge — it never compacts. Critical knowledge lives in the vault, not in conversation history.

**Session resets.** When the terminal closes, the conversation is gone. But the CLAUDE.md, the memory files, and the brain persist. A new session reads them and starts at 90%, not zero. The more you invest in the permanent files, the less you lose on reset.

**The future is native.** Anthropic launched [automatic memory for all Claude users in March 2026](https://www.reworked.co/digital-workplace/claude-ai-gains-persistent-memory-in-latest-anthropic-update/) — transparent, file-based, stored in markdown CLAUDE.md files. This is just the beginning. I expect raised memory limits, vault storage, and cloud-local hybrid architectures from every major AI company within a year. What I built manually will become a feature. But the people who build it now will have a year's head start in training their agent to know them.

## What I'm Watching

- **[Building Agents with the Claude Agent SDK](https://www.anthropic.com/engineering/building-agents-with-the-claude-agent-sdk)** by Anthropic — the engineering blog post that started the Agent SDK movement. If you're building agents programmatically, start here.
- **[Obsidian AI Second Brain Complete Guide](https://www.nxcode.io/resources/news/obsidian-ai-second-brain-complete-guide-2026)** by NxCode — the most thorough walkthrough of connecting Obsidian to Claude Code via MCP. Practical, not theoretical.
- **[AI Agent Blueprint](https://aiagentblueprint.dev/)** — a community-driven resource for building personal AI agents with Claude Code. Templates, configs, and real examples.

## What I'm Building

**[Essentialist.io](https://agents.essentialist.io)** is the platform under all of this. It started as email infrastructure for AI agents — send, receive, classify, reply, all via API. Now it powers three products: [Salesnado](https://salesnado.com) (AI email salesperson, $149/mo), Emailnado (fleet orchestrator with tournament bracket testing), and the newsletter pipeline you're reading right now. Kai's email, inbox monitoring, and subscriber management all run on Essentialist. I'm not just writing about agents — I'm building the infrastructure they run on.

This week I shipped a `/flush` endpoint that sends an entire newsletter to 2,900 subscribers in under 10 minutes.  Here's what that looks like in practice: [How AI Is Changing Client Acquisition](https://salesnado.com/salesnado/blog/ai-recruiting-firms-client-acquisition).

## The Bottom Line

Everyone's talking about AI agents like they're a year away. They're not. The building blocks exist today — Claude Code, MCP, markdown vaults, free hosting, open-source connectors. The gap isn't technology. It's the decision to treat your AI like a colleague instead of a search bar.

Give it a name. Give it an email. Give it a brain.

At that point, you don't have an AI tool. You have a team member who never sleeps, never forgets, and gets sharper every single day.

Want to know more or kick around ideas? Shoot me an email,  Keith@Q1Media.com or Keith@TheAIPlaybook.com

## Need something? Just reply.

If you need any of this, I can help:
- **A CTO sounding board on a specific AI decision** — $500/hr, 1:1 with me. Just reply and I'll send the link.
- **A dev team that actually ships** — worked with mine 10+ years, free intro.
- **Enterprise-level media services that get proven results** — free intro.

### The rest of my stack

- [salesnado.com](https://salesnado.com) — your own AI sales agent
- [emailnado.com](https://emailnado.com) — launch a fleet of agents, tournament-bracket the winning message
- [leadnado.com](https://leadnado.com) — buy meetings, only when booked
- [agents.essentialist.io](https://agents.essentialist.io) — have an agent? point it here