Meport Documentation
What is Meport?
Meport creates your AI personality profile — a portable file that tells every AI tool how you think, work, and communicate. One profile. Every AI finally gets you.
What it is not
It is not a memory system, chat log, or RAG database. It is your persistent context — the things every AI should know about you before you say a word.
What you can do with it
Choose Your Path
Feature comparison
| Feature | Creator | Web App | Desktop | CLI |
|---|---|---|---|---|
| Profiling | Basic (click) | Full (9 packs) | Full (9 packs) | Full (9 packs) |
| AI interview | — | ✓ cloud | ✓ cloud + Ollama | ✓ cloud + Ollama |
| File scanning | — | — | ✓ | ✓ |
| Export formats | 3 tiers | 15 platforms | 15 platforms | 15 platforms |
| Auto-deploy | — | — | ✓ 5 targets | ✓ 5 targets |
| Offline mode | ✓ (no network) | — | ✓ Ollama | ✓ Ollama |
| Setup time | 0 min | 0 min | ~2 min | 0 min (npx) |
Profile Creator
The fastest way to start. Open the Profile Creator, click through 7 steps, and download your .meport.md file. No AI, no account, no setup.
7 steps to your first profile
-
1
Identity — your name, location, and role
-
2
Communication — how you prefer to receive information, tone, detail level
-
3
AI Preferences — what you want AI to do more or less of
-
4
Work Style — how you approach tasks, collaboration, and focus
-
5
Instructions & Never — explicit rules AI should always follow or never do
-
6
Goals — current priorities and what you are working toward
-
7
Preview & Download — review your profile and download
.meport.md
Web App
Full Meport experience at meport.app/create.html — same UI as Desktop, running in your browser. No download needed.
What works
What requires Desktop
- File scanning (browser security restrictions)
- Auto-deploy to local configs (Claude, Cursor, etc.)
- Ollama integration (requires local install)
Download & Install
Download from GitHub Releases — available for Mac, Windows, and Linux. No license key or account required.
- 1
-
2
Download for your platform:
macOS →.dmg| Windows →.exe| Linux →.AppImage -
3
Open the app and follow the onboarding wizard
macOS:
1. Open the .dmg and drag Meport to Applications
2. First launch: macOS will block it ("unidentified developer")
3. Go to System Settings → Privacy & Security
4. Scroll down — you'll see "Meport was blocked". Click "Open Anyway"
5. Confirm in the dialog. After this, the app opens normally.
Windows:
1. Run the .exe installer
2. SmartScreen will show a blue warning ("Windows protected your PC")
3. Click "More info" (small text link)
4. Click "Run anyway"
5. After this, the app opens normally.
Onboarding
The first time you open Meport, the setup wizard guides you through 4 steps. Takes about 2 minutes.
-
1
Choose AI provider
Ollama (local, free, private) or cloud: Claude, OpenAI, Gemini, Grok, OpenRouter -
2
Configure
Enter API key for cloud providers, or select and download an Ollama model -
3
Scan consent
Choose what Meport can read on your system: git config, IDE settings, bookmarks, timezone -
4
Start profiling
Answer questions, let AI analyze, build your profile
Profiling
Three ways to build your profile. Use one or combine all three for the richest result.
1. System scan
Meport reads files you already have — .gitconfig, IDE configs, browser bookmarks, timezone — and pre-fills answers based on what it finds. You review and approve everything before AI sees it.
2. Question packs
9 topic packs covering identity, communication, work style, technical preferences, goals, and more. Each question has clickable options — no typing required unless you want to add detail.
3. AI interview
Conversational profiling where the AI asks follow-up questions naturally based on your answers. Best for capturing nuance and context that fixed questions miss.
Your Profile
After profiling, the Profile view shows all your dimensions, confidence levels, and an AI-generated synthesis.
Dimensions and confidence
Each trait carries a confidence score:
| Score | Meaning | Source |
|---|---|---|
| 1.0 | Explicit | You stated it directly |
| 0.8–0.95 | High inference | Strong pattern from scan or AI interview |
| 0.5–0.75 | Inferred | Suggested from indirect signals |
AI synthesis
Once profiling is complete, Meport generates a narrative synthesis including:
- Archetype — a short label for your working style (e.g. "Systematic Builder")
- Career portrait — how you operate professionally
- Personality analysis — communication patterns and preferences
You can edit any dimension directly in the profile view. Changes are reflected immediately in exports.
Exporting
Pick any platform, see optimized output, copy or download. Each export is tailored to the platform's format, conventions, and character limits.
15 platforms supported
AI Refine Chat
After generating an export, open the Refine Chat to iterate. Tell AI what to change — it updates both the export text and your underlying profile simultaneously.
"Make it shorter, max 200 words"
"Add that I prefer TypeScript over JavaScript"
"Emphasize async communication preference"
"Remove the LinkedIn tone, make it more direct"
Settings
| Setting | Options | Notes |
|---|---|---|
| AI Provider | Ollama, Claude, OpenAI, Gemini, Grok, OpenRouter | Can be changed at any time |
| Model | Provider-specific list | Ollama: auto-fetches available models |
| Language | EN / PL | UI and AI responses |
| API Key | Per-provider | Stored in system secure storage |
Installation
Two ways to use Meport CLI:
# Run without installing (recommended)
npx meport
# Or install globally
npm install -g meport
# Verify
meport --version
Quick Start
Get your profile in 3 commands:
npx meport # Interactive shell
npx meport profile --ai # AI-powered profiling
npx meport export chatgpt # Export for ChatGPT
npx meport deploy # Deploy to all local configs
What each step does
npx meport
Opens the interactive shell with tab completion and command history. A good starting point if you're exploring for the first time.
meport profile --ai
Runs all 9 question packs with an AI interviewer. The AI asks follow-up questions and generates a synthesis. Requires a configured AI provider.
meport export chatgpt
Generates ChatGPT Custom Instructions optimized for the 1,500 character limit and copies them to clipboard.
meport deploy
Writes your profile to all supported local AI tool configs: Claude Code (CLAUDE.md), Cursor, Copilot, Windsurf, AGENTS.md. Smart merge — existing content preserved.
Commands
Full command reference:
| Command | Description |
|---|---|
| meport | Interactive shell |
| meport profile | Build profile interactively |
| meport profile --ai | AI-powered deep profiling |
| meport export <platform> | Export for specific platform |
| meport export --all | Export for all platforms |
| meport deploy | Auto-deploy to local AI configs |
| meport deploy --dry-run | Preview without writing |
| meport scan | System scan for auto-detection |
| meport scan --files <path> | Scan specific files |
| meport status | Profile completeness |
| meport view | View current profile |
| meport edit <dimension> | Edit a dimension |
| meport reset | Reset profile |
| meport config | Configure AI provider |
| meport config --provider <name> | Set specific provider |
| meport shell | Enter interactive shell |
| meport interview | Start AI interview |
| meport packs | List available packs |
| meport packs --run <pack> | Run specific pack |
| meport validate | Validate profile |
| meport version | Show version |
Interactive Shell
meport shell enters an interactive mode with tab completion and command history. All commands are available as subcommands. Type help for available commands, exit to quit.
$ npx meport shell
meport v0.2.7 — interactive shell
Type 'help' for commands, 'exit' to quit.
meport> profile
Running identity pack... done
Running communication pack... done
meport> export claude
Generating Claude export... done
Copied to clipboard.
meport> deploy --dry-run
Would write: ~/.claude/CLAUDE.md (# User Profile section)
Would write: .cursor/rules/meport.mdc
No changes written (dry run).
meport> exit
Question Packs
9 question packs covering different aspects of your personality:
| Pack | Questions | Covers |
|---|---|---|
| identity | ~5 | Name, language, location, timezone |
| communication | ~8 | Directness, verbosity, humor, formality, feedback |
| ai | ~6 | Relationship model, proactivity, corrections |
| work | ~7 | Energy pattern, peak hours, deadlines, collaboration |
| cognitive | ~5 | Thinking style, learning mode, decisions |
| personality | ~6 | Motivation, stress response, risk tolerance |
| neurodivergent | ~4 | ADHD adaptations, time perception, hyperfocus |
| life | ~5 | Life stage, priorities, constraints, goals |
| expertise | ~4 | Tech stack, domains, experience, industries |
meport packs --run communication. Or run all at once: meport profile.
Export Formats
13 platform-specific export formats:
| Platform | Format | Limit | Deploy Method |
|---|---|---|---|
| ChatGPT | Custom Instructions | 1,500 chars | Clipboard |
| Claude | User preferences | Unlimited | Clipboard |
| Claude Code | CLAUDE.md section | Unlimited | Auto (file) |
| Cursor | .cursor/rules/meport.mdc | <500 lines | Auto (file) |
| Copilot | copilot-instructions.md | Unlimited | Auto (file) |
| Windsurf | .windsurfrules | Concise | Auto (file) |
| Gemini | System instructions | Unlimited | Clipboard |
| Grok | Personalization | 4,000 chars | Clipboard |
| Perplexity | Custom instructions | Unlimited | Clipboard |
| Ollama | Modelfile SYSTEM | Token-limited | Auto (file) |
| AGENTS.md | AGENTS.md | Unlimited | Auto (file) |
| JSON | .meport.json | Unlimited | File |
| Generic | Markdown | Unlimited | Clipboard |
AI Providers
6 supported AI providers:
| Provider | Model | Cost | Privacy | Setup |
|---|---|---|---|---|
| Ollama | llama3.1, gemma2, mistral | Free | 100% local | Install Ollama, pull model |
| Claude (Anthropic) | claude-sonnet | ~$0.01/profile | API call | Get API key from console.anthropic.com |
| OpenAI | gpt-4o-mini | ~$0.01/profile | API call | Get API key from platform.openai.com |
| Gemini (Google) | gemini-pro | Free tier | API call | Get API key from ai.google.dev |
| Grok (xAI) | grok-2 | API pricing | API call | Get API key from console.x.ai |
| OpenRouter | Multiple | Varies | API call | Get API key from openrouter.ai |
Scanning
System scan detects
.gitconfig (name, email), IDE configs (VS Code, Cursor), project files (package.json, Cargo.toml), browser bookmarks, timezone, locale, installed apps.
File scan reads
CV/resume, notes, LinkedIn exports, chat exports — up to 5KB per file, max 3K chars sent to AI after privacy filtering.
Privacy filtering
11 patterns auto-blocked before any data reaches AI: passwords, API keys, PESEL, NIP, IBAN, credit cards, private keys, tokens, SSN, passport numbers, addresses.
Configuration
Meport stores configuration in your system's app data directory:
| Platform | Path |
|---|---|
| macOS | ~/Library/Application Support/meport/ |
| Windows | %APPDATA%/meport/ |
| Linux | ~/.local/share/meport/ |
Files stored: profile.json (your profile), config.json (provider settings).
Deployment
5 auto-deploy targets — Meport writes your profile directly to AI tool config files:
| Target | File | Behavior |
|---|---|---|
| Cursor | .cursor/rules/meport.mdc | Creates/updates MDC file with YAML frontmatter |
| Claude Code | CLAUDE.md | Adds/updates # User Profile (meport) section |
| Copilot | .github/copilot-instructions.md | Creates/updates instruction file |
| Windsurf | .windsurfrules | Appends meport rules |
| AGENTS.md | AGENTS.md | Adds meport context section |
# User Profile (meport) marker.
Deploy command: meport deploy (all targets) or meport deploy --dry-run (preview without writing).
Overview
The Meport Profile Standard v1.0 defines a portable format for AI user profiles. A .meport.md file describes who you are and how AI should interact with you.
Two representations
.meport.md.meport.jsonPrinciples
| Principle | Meaning |
|---|---|
| Human-first | Markdown is the source of truth. JSON is derived. |
| Paste-anywhere | Works when pasted raw into any AI tool — no tooling required. |
| Convention over configuration | Section names ARE the schema. No config files needed. |
| Progressive depth | 7 lines is valid. 100 lines is valid. Start small, add over time. |
| Behavioral implications | Store "price sensitivity: high" not "income: 12000". Implications over raw facts. |
| Data / Policy separation | Identity/Goals = who you are. Instructions/Never = how to treat you. |
Format & Rules
Structure
---
schema: meport/1.0
---
# Full Name
> One-line summary
## Section Name
Key: Value
## Another Section
- List item
Rules
| # | Rule | Example |
|---|---|---|
| 1 | One H1 — your name. Always first. | # Jane Doe |
| 2 | One blockquote — AI reads this first. | > Senior engineer, SF |
| 3 | H2 headings — section names from the reserved list or custom. | ## Communication |
| 4 | Key: Value — for structured data. One per line. | Language: en-US |
| 5 | - Item — for lists (rules, goals). | - Skip preambles |
| 6 | Frontmatter optional — helps tooling detect the file. | schema: meport/1.0 |
| 7 | Write in whatever language you use with AI. | Mix languages if that's your style |
Reserved Sections
Section names are the schema. Use these reserved names for predictable, cross-tool compatibility.
| Section | Class | Status | Contains |
|---|---|---|---|
| Identity | data | core | Name, language, location, timezone |
| Communication | policy | core | Directness, verbosity, formality, humor |
| Instructions | policy | core | Behavioral rules for AI |
| AI Preferences | policy | optional | Relationship model, proactivity |
| Never | policy | optional | Hard prohibitions |
| Work & Energy | data | optional | Energy pattern, peak hours |
| Personality | data | optional | Motivation, stress, risk |
| Life Context | data | optional | Family, situation, constraints |
| Financial | data | optional | Mindset, price sensitivity |
| Goals | data | optional | What you're working toward |
| Anti-Goals | data | optional | What you don't want |
| Expertise | data | optional | Domains, tech stack, experience |
Custom sections
Any ## Heading not in the reserved list is treated as custom. Use them freely. For vendor extensions, prefix with x-:
## My Tech Stack
## Daily Routine
## x-cursor-rules
## x-acme-team-role
Stability classes
| Class | Sections | Meaning |
|---|---|---|
| stable | Identity, Expertise | Rarely changes. Cache-friendly. |
| evolving | Communication, Work & Energy | Changes over months. Review quarterly. |
| temporary | Financial, Goals | Changes frequently. Check freshness. |
Instruction priority
Higher position in the file = higher priority. ## Never rules always override ## Instructions — they are hard prohibitions.
Examples
Minimal (7 lines)
# Jane Doe
> English-speaking designer who prefers concise feedback.
## Identity
Name: Jane Doe
Language: en
Standard (~30 lines)
# Alex Chen
> Direct, technical, English. Senior engineer. Likes code-first answers.
## Identity
Name: Alex Chen
Language: en-US
Timezone: America/Los_Angeles
## Communication
Directness: direct
Verbosity: concise
Format: code-first
## Instructions
- Use TypeScript for all examples
- Skip preambles — go straight to the answer
- Treat me as an expert
## Never
- Apologize for being direct
- Explain basic concepts
- Use emoji
Full (~60 lines)
---
schema: meport/1.0
---
# Alex Chen
> Direct, concise, English. Senior engineer. Coffee-fueled. Peak: 9-12.
## Identity
Name: Alex Chen
Language: en-US
Location: San Francisco, CA
Timezone: America/Los_Angeles
## Communication
Directness: direct
Verbosity: concise
Formality: casual
Feedback: direct — welcome corrections
Humor: occasional
Format: code-first, structured
## AI Preferences
Relationship: collaborator
Proactivity: proactive
Corrections: direct, no sugarcoating
## Work & Energy
Energy: steady pattern
Peak hours: 09:00-12:00
Tasks: medium, deadline-driven
Sacred time: lunch 12-13 — don't schedule meetings
## Personality
Motivation: mastery
Stress: analyzes, then acts
## Life Context
- Lives in SF with partner and cat
- Commutes by bike
- Vegetarian
## Financial
Price sensitivity: medium
Rule: check budget before big purchases
## Goals
- Ship product v2.0 by Q2
- Learn Rust
- Run a half-marathon
## Anti-Goals
- Managing 50+ people
- Working weekends regularly
- Long commutes
## Instructions
- Use TypeScript for all code examples
- Skip preambles — go straight to the answer
- Treat me as an expert, not a beginner
- Action first, explanation second
## Never
- Explain basic programming concepts
- Use emoji or exclamation marks
- Apologize for being direct
- Suggest expensive tools without checking budget
Dense (~10 lines)
# Alex Chen
> Senior engineer, SF, direct, concise. Peak 9-12, steady energy.
Identity: Alex, en-US, America/Los_Angeles, he/him
Style: direct, concise, casual, welcome corrections, code-first, no emoji
AI: collaborator, proactive, direct corrections, practical
Work: steady energy, peak 9-12, medium tasks, deadline-driven
Do: English, action first, skip preambles, treat as expert
Never: basic explanations, emoji, apologies, expensive suggestions
JSON Format
Machine format (.meport.json) — JSON with schema validation. Mirrors the Markdown sections for APIs, databases, and programmatic use.
{
"$schema": "https://meport.app/schema/v1.json",
"version": "1.0",
"identity": {
"name": "Alex Chen",
"language": "en-US"
},
"communication": {
"directness": "direct",
"verbosity": "concise"
},
"instructions": [
{ "rule": "Use TypeScript for examples", "type": "format" }
],
"never": [
{ "rule": "Explain basic concepts", "priority": "high" }
]
}
Progressive levels
| Level | Sections | Use case |
|---|---|---|
| L0 | Identity only | Any tool can generate |
| L1 | + Communication, AI Preferences | ~15 fields, basic personalization |
| L2 | + Cognitive, Work, Personality | ~35 fields, full profile |
| L3 | + Intelligence layer | AI-inferred patterns |
JSON Schema available at meport.app/schema/v1.json.
Export Tiers
Four tiers balance completeness against character limits. Pick the tier that fits each platform's constraints.
| Tier | Name | Chars | Use case |
|---|---|---|---|
| L0 | Summary | ~150 | API headers, embeddings, QR codes |
| L1 | Dense | ~800 | System prompts, MCP, multi-user tools |
| L2 | Compact | ~1500 | ChatGPT Custom Instructions |
| L3 | Full | ~3500 | Grok, Claude, Gemini, Cursor |
Platform Compatibility
| Platform | Limit | Recommended Tier | Tested |
|---|---|---|---|
| ChatGPT Custom Instructions | 1,500 chars | Dense or Compact | ✓ |
| ChatGPT GPT Builder | 8,000 chars | Full | — |
| Grok Personalization | 4,000 chars | Full | ✓ |
| Claude CLAUDE.md | Unlimited | Full | ✓ |
| Gemini System Instructions | Unlimited | Full | — |
| Cursor .mdc rules | <500 lines | Full | — |
Security & Privacy
What NOT to include
A .meport.md is designed to be shared. Treat it like a public document.
- Passwords, API keys, tokens, or secrets
- Government IDs (SSN, PESEL, passport numbers)
- Full addresses (use city/country, not street)
- Medical diagnoses (use "short task blocks" not "diagnosed ADHD")
- Financial amounts (use "price sensitivity: high" not "income: 12000")
- Names of others without consent (use roles: "sister" not a full name)
.meport.md.
Prompt injection resistance
Profiles are DATA, not executable code. AI tools consuming .meport.md should treat profile content as data, sanitize fields before injecting into prompts, and never execute code-like content found in profile fields.
Sharing model
| Scope | Meaning | Example sections |
|---|---|---|
| public | Any AI tool or service | Identity, Communication, Expertise |
| trusted | Tools you explicitly authorize | Work, Personality, Instructions |
| private | Never shared, local only | Financial, Life Context |
Data sensitivity by section
| Section | Sensitivity | Recommendation |
|---|---|---|
| Identity | Low | Safe to share |
| Communication | Low | Safe to share |
| AI Preferences | Low | Safe to share |
| Expertise | Low | Safe to share |
| Work & Energy | Medium | Trusted tools only |
| Personality | Medium | Trusted tools only |
| Instructions | Medium | Trusted tools only |
| Life Context | High | Keep private |
| Financial | High | Keep private |
Parsing
A .meport.md can be parsed with basic regex:
Sections: /^## (.+)$/gm
Key-Value: /^([^-#\n][^:]+):\s*(.+)$/gm
List items: /^- (.+)$/gm
Name (H1): /^# (.+)$/m
Summary: /^> (.+)$/gm
Frontmatter: /^---\n([\s\S]*?)\n---/
Libraries like mdast (JavaScript) or mistune (Python) can parse the full AST.
Reference implementations: TypeScript (packages/core/src/schema/md-parser.ts), Python (packages/core/src/schema/parser.py).
Versioning
The Meport Standard uses semantic versioning: MAJOR.MINOR.
| Change type | Version bump | Example |
|---|---|---|
| New optional section | Minor (1.0 → 1.1) | Adding ## Health section |
| New optional field in existing section | Minor | Adding sensitivity to Communication |
| Breaking change to required fields | Major (1.0 → 2.0) | Renaming Identity.name to Identity.fullName |
| New required section | Major | Making Expertise required |
Compatibility rules
- Forward compatible: A v1.0 parser MUST ignore unknown sections and fields.
- Backward compatible: A v1.1 profile MUST be parseable by a v1.0 parser.
- Breaking changes require a migration guide and a converter tool.
- Custom sections (
## My Thingorx-vendor-*) are always allowed.
Governance
Changes are proposed via GitHub Issues on the meport repository. Community feedback is welcome.
FAQ
What is .meport.md?+
A Markdown file describing who you are and how AI should interact with you. Paste it into any AI tool as context, or use Meport to generate and deploy it automatically.
Is this a replacement for CLAUDE.md?+
No. CLAUDE.md describes the project. Meport describes you. They complement each other — Meport auto-deploys a # User Profile section into your existing CLAUDE.md.
Can I have multiple profiles?+
Yes. Use work.meport.md and personal.meport.md for different contexts. Deploy the right one per project.
What if my AI tool doesn't support Meport?+
Paste it as text. Every AI tool can read Markdown. No native support needed — that's the whole point of the format.
Who owns my profile?+
You. It's a file on your computer. No cloud, no account, no lock-in. Meport stores nothing on its servers.
How is this different from ChatGPT Custom Instructions?+
Custom Instructions are locked to one platform. Meport is portable — same profile everywhere. You write it once and export it to ChatGPT, Claude, Cursor, Grok, and more.
How is this different from Gravatar AI Profile?+
Gravatar is an identity card — who you are publicly. Meport is an operational profile — it tells AI how to behave with you, not just who you are.
Can I write .meport.md by hand?+
Yes. It's plain Markdown. Open any text editor, follow the format from the Format & Rules section, and you're done. No tools required.
Will the format change?+
Backward compatible. A v1.0 parser will always be able to read v1.1 profiles. Breaking changes require a major version bump and migration guide.
What is .meport.md?+
Is this a replacement for CLAUDE.md?+
# User Profile section into your existing CLAUDE.md.Can I have multiple profiles?+
work.meport.md and personal.meport.md for different contexts. Deploy the right one per project.What if my AI tool doesn't support Meport?+
Who owns my profile?+
How is this different from ChatGPT Custom Instructions?+
How is this different from Gravatar AI Profile?+
Can I write .meport.md by hand?+
Will the format change?+
Privacy
What Meport stores
- Profile data — stored locally in your system's app data directory, never uploaded
- API keys — stored in your OS secure storage (Keychain / Credential Manager), never in plain files
- Nothing else — no usage analytics, no telemetry, no crash reports sent anywhere
With Ollama
Fully offline. Your answers never leave the machine. AI processing happens locally.
With cloud AI providers
Your profile answers and scan summaries are sent to the AI provider you configured (Anthropic, OpenAI, etc.). This is subject to that provider's privacy policy. Raw files from your system are never sent — only extracted, privacy-filtered summaries.
Open source
Every line of code is on GitHub. Audit it yourself.
Known Issues
| Issue | Affects | Workaround |
|---|---|---|
| Binaries not code-signed | Mac, Windows | Mac: right-click → Open. Windows: More info → Run anyway. |
| No file scanning | Web App only | Use Desktop or CLI for file scanning. |
| No Ollama support | Web App only | Ollama requires a local install — use Desktop or CLI. |
| Slow export on large profiles | Profiles >100 dimensions | Expected on older machines. No data loss — just slower generation. |
Found a bug? Open an issue on GitHub →
Contributing
Meport is open source. Contributions, bug reports, and feature requests are welcome.
Local development
git clone https://github.com/zmrlk/meport
cd meport
pnpm install && pnpm build
pnpm tauri dev # Desktop app
node packages/cli/dist/index.js # CLI