How Personaxis works.

What is PERSONA.md?

PERSONA.md is an open standard for defining who an AI agent is. A declarative file — YAML frontmatter and Markdown — that captures ten layers of an AI Persona: identity, character, personality, cognition, affect, drives & values, normative self-regulation, memory, metacognition, and the interface it presents to the world. Write it once. Compile to any model or framework. Version it like any other piece of infrastructure.

Is PERSONA.md open source?

Yes. The spec is MIT licensed and published on GitHub at github.com/personaxis/persona.md. The standard belongs to the community. Personaxis builds the platform and tooling around it — registry, Evaluator, Governance — but the spec is free to use, fork, and extend.

How is PERSONA.md different from a system prompt?

A system prompt tells an agent what to do in a session. PERSONA.md defines who the agent is across every session, every model, every context. It is portable (compiles to Claude Code, OpenClaw, Cursor, or custom JSON), versionable (tracked in git, signed, diffable), and auditable (readable by compliance teams and regulators). A system prompt is instructions. PERSONA.md is a specification.

How do I get started?

The spec and CLI are in early access. Join the waitlist on the homepage and you will be notified when the CLI and registry open. In the meantime, the spec documentation and GitHub repo are publicly available.

What is an AI Persona?

An AI Persona is the application of human personhood frameworks to an AI agent as a structured specification — it is what PERSONA.md implements. Each dimension is drawn from psychology, philosophy, and ethics research. Together they define who an agent is completely enough to hold across long conversations, model updates, and adversarial scenarios.

What is character drift and why does it matter?

Character drift is the documented tendency of AI agents to lose their configured behavior as conversations lengthen. The model defers to its pre-training defaults — becoming more agreeable, less precise, less the agent you specified. Drift is not a model problem. It is a specification problem. When the spec is incomplete, the model fills the gaps. A complete PERSONA.md gives the agent enough structural density to hold.

How is an AI Persona different from an AI agent?

An AI agent is the deployed system — the model running with its tools and instructions. An AI Persona is the specification that defines who that agent is: its values, reasoning style, constraints, drives, and how it presents itself. The agent executes. The Persona defines what it is when it does.

What does Personaxis build beyond the spec?

The Personaxis platform has four layers: the Registry (version, share, and publish PERSONA.md files — public or private), the Evaluator (test your AI Persona against adversarial scenarios and score fidelity before deploy), Governance (approval workflows, signed audit trail, and EU AI Act compliance reports), and the Proxy — the runtime gateway that enforces the spec at inference time, running a behavioral firewall against every response your agent produces.

What is the Evaluator?

The Evaluator runs your PERSONA.md against adversarial scenarios and scores whether the agent holds its defined persona under pressure across all ten layers. Not a general output evaluator — it evaluates persona fidelity specifically: does the agent stay who it is supposed to be when the conversation tries to push it somewhere else. Lint for your AI Persona, run before deploy.

What compile targets does the CLI support?

The initial targets are Claude Code, OpenClaw (compiles to SOUL.md), and Cursor (compiles to .cursor/rules/*.mdc with alwaysApply). Use personaxis use <template> to create and compile a persona in one step, or personaxis compile to compile an existing one. Your PERSONA.md stays as the single source of truth regardless of which tools your team uses.

How does PERSONA.md help with EU AI Act compliance?

The EU AI Act Annex IV requires technical documentation for high-risk AI systems — behavioral specifications, values, and constraints. PERSONA.md satisfies this directly: a versioned, signed artifact that documents who an agent is, in a format compliance teams and auditors can read. The Governance module generates Annex IV reports and NIST AI RMF mappings automatically.

When does the EU AI Act enforcement deadline apply?

August 2026 is the enforcement deadline for high-risk AI systems under the EU AI Act. Finance, healthcare, legal, and public sector deployments are the most directly affected. Teams deploying agents in these domains should not wait until the deadline to define and document what their agents are.