Me.txt – A personal identity file for AI agents
Standard markdown profile—solves a real friction point, but spec adoption is the entire bet.

Finally solves locked-in custom instructions with a portable robots.txt for people.
AI power users, developers building AI tools
llms.txt · humans.txt · robots.txt
Every AI tool has some version of "custom instructions" but they're locked inside that tool. Switch from ChatGPT to Claude to Gemini and you're starting from scratch each time. Your voice, your preferences, your expertise, all gone.
identity.txt is a plain text markdown file (same philosophy as llms.txt, humans.txt, robots.txt) that captures who you are for AI tools. Drop it at yourdomain.com/identity.txt or paste it into any context window.
The spec is deliberately minimal:
- H1 with your name - H2 sections for Voice, Expertise, Background, Preferences, Terms - A Terms section with machine-readable consent (open, attribution, prompt-only, restricted, none)
The Terms section is interesting, maybe. It's like robots.txt for personal identity. A social contract, not a legal one. We think consent and identity overlap in ways worth exploring. We def don't have the answers now. But open to the questions.We're also experimenting with hosted identity files at identitytxt.org. You authenticate with Google, pick a handle, and get a permanent URL. But self-hosting is the primary use case. The hosted version is just convenience. Maybe. We don't know.
Look, this is early and experimental. We built it because we kept pasting the same context into every AI conversation and thought there should be a convention for it. The spec is CC-BY 4.0 and on GitHub: https://github.com/Fifty-Five-and-Five/identitytxt
We're genuinely interested in whether this resonates or if the problem gets solved differently. What are we missing?
Standard markdown profile—solves a real friction point, but spec adoption is the entire bet.
Portable .pid identity format lets AI minds migrate across models and hardware.
Incremental AST parsing beats full re-renders for long LLM markdown streams.
LLM-generated Markdown with embedded forms and approval gates instead of prose instructions.
Evidence-mapped CV beats PDF for AI recruiter parsing, but applies only to ATS that read these formats.
Validates llms.txt and AI robot rules before AI crawlers ignore your content.