maiiamaiia
GitHub
AI Human Passport · v0.1·Open source

Stop re-explaining
yourself to every AI.

One portable Passport that Claude, Cursor, ChatGPT, and Gemini all read from. Fills itself as you chat. Shows you your AI day with mood and per-model attribution.

Get started Open dashboard
How it works

Three steps. Then it learns.

01

Connect any AI

One scoped token, four configs ready to paste — Claude Desktop, Cursor, ChatGPT Custom GPT, or any HTTP client.

02

Chat normally

The agent suggests profile updates as you talk. Low-stakes facts auto-approve. Sensitive ones never get suggested.

03

See your day

Mood arc + per-model frustration. Monthly Recap tells you which model was MVP and which one drove you up the wall.

What you'll see

Not slides. The actual product.

Here's what each page looks like once you've chatted for a few days.

Today

A daily mood arc with model attribution.

Who frustrated you, who carried the day, and a couple of anecdotes. Opt-in. Ephemeral by default. Never stores your conversation text.

  • Stress + focus + satisfaction scores
  • Per-model "rage" and "MVP" cards
  • Playful or neutral anecdotes — your call
maiia.dev/today
Today
Tuesday, June 2
Your day, in one line

Deep focus on agent infrastructure. Opus carried the architecture work; GPT-5 ate four threads.

focusedfrustratedcurious
Mood arc
68%
Most rage
openai
/gpt-5
4×
MVP
anthropic
/opus-4.7
6×
14:18

Your fourth rephrase of "no, the OTHER kind of MCP" suggests GPT-5 needs a vacation. Or you do.

Recap

A monthly recap. Yes, model awards.

Your archetype, top themes, and — most importantly — your MVP model and your most rage-inducing model.

  • Archetype (Systems Builder, Strategic Operator, …)
  • Model awards: MVP, Most rage-inducing, Workhorse
  • A portable context card to paste anywhere
maiia.dev/recap
AI Recap
May
You used AI mostly as
The Systems
Builder

You move fast from idea → architecture → MVP. Trusted Opus with high-stakes work, mostly satisfied.

MVP of May
anthropic
/opus-4.7
Most rage
openai
/gpt-5
Top themes
  1. 01MCP
  2. 02AI agents
  3. 03personal context
  4. 04startup architecture
Profile

What every AI reads about you, first.

Layered: facts you provided, preferences inferred, all clearly labeled. Edit anything. Delete anything. The agent's reasoning lives next to the claim.

  • Color-coded by category (work, communication, identity, …)
  • Source badge: "You" vs "AI · 87% confident"
  • Sensitive categories off by default, always
maiia.dev/profile
Factual profile
What we know.
Passport summary

A product-minded AI power user focused on agent infrastructure, MCP, and digital identity. Direct, concrete, implementation-first.

All · 12Work · 4Communication · 3Preference · 2
  • work
    Stack

    TypeScript + Drizzle + Hono + Next.js. Prefers monorepos.

    You
  • communication
    Tone

    Direct, practical, strategic. No long preambles.

    You
  • preference
    Implementation-first

    Show MVP scope, architecture, tradeoffs — skip the framing.

    AI · 87%
  • identity
    Role

    Founder / product strategist

    You
Approvals

Agents suggest. You approve.

When an agent notices something stable about you, it lands in this inbox with evidence and confidence. One click to keep, one to reject. Or set thresholds to auto-approve low-stakes facts.

  • Per-model attribution — see which agent suggested what
  • Auto-approve thresholds by category
  • Sensitive-topic filter prevents inappropriate suggestions
maiia.dev/approvals
Approvals
3 to review.
  • preference92%anthropic/opus-4.7
    Prefers TypeScript over Go

    For backend services. Cited TS ergonomics in three conversations this week.

  • work88%anthropic/opus-4.7
    Building agent infrastructure

    MCP, scoped tokens, observer processes — multiple deep-dive sessions.

  • communication76%openai/gpt-5
    Wants tradeoffs explicit

    "Pros and cons" appears in 6+ replies this week — even on small choices.

Sharing

One token. Four platforms. Revoke anytime.

Each agent gets its own scoped JWT. Pre-baked configs for Claude/Cursor (MCP), ChatGPT (Custom GPT Actions), Gemini (Function Calling), and plain REST. Every call is logged.

  • Granular scopes: basic profile, comms, daily pulse, suggest_updates …
  • Audit log of every read/write
  • Public OpenAPI spec for Custom GPTs and SDK clients
maiia.dev/sharing
Sharing & MCP
Who can see what.
  • Claude Desktop
    anthropic/opus-4.7
    4 scopes
  • ChatGPT (work)
    openai/gpt-5
    4 scopes
  • Cursor
    anthropic/sonnet-4.6
    4 scopes
Generated token — shown once
phk_eyJhbGciOiJIUzI1NiJ9.eyJzdWIi…fX9.x9k2
basic_profilecomm_styledaily_pulsesuggest_updates
The trust model

You own this. We mean it.

Sensitive categories off by default

Health, finances, religion, politics, sex, mental health — never inferred or stored unless you flip them on yourself.

Agents suggest, never approve

Every AI-inferred profile item starts as a suggestion. Auto-approve only what you allow, only above the confidence threshold you set.

Raw conversation text isn't kept

The Pulse Layer stores summaries, not transcripts. Evidence is paraphrased. Sensitive content is dropped before it touches the database.

Export. Revoke. Delete.

JSON or Markdown export, anytime. Revoke an agent's access in one click. Account deletion cascades to every row.

Open source — verify what we do

Code lives at github.com/headlessButSmart/maiia. Audit it, fork it, self-host it.

Portable forever

JSON export is the canonical format. If you leave, you take your Passport with you. No lock-in.

One Passport. Every agent. No re-onboarding.

Claude Desktop, ChatGPT, Cursor, Gemini, custom agents — they all read from the same MCP/REST endpoint. When you teach one agent something, every other agent benefits.

Get started Read the source