The AI File Standard for Expert Knowledge
An XML-based structured file format that packages expertise from lawyers, doctors, and consultants into a form that AI can execute immediately. Multi-LLM compatible — works identically on ChatGPT, Claude, and Gemini.
Six Reasons .xagt is Different
Most AI prompts are fragile, stateless, and disposable. .xagt is designed to be permanent, portable, and protected — the first truly deployable AI knowledge asset.
Deterministic Execution
Embed Python code directly in <xcd> for calculations, lookups, and logic that cannot hallucinate. Tax computations, legal clause parsing, medical dosage checks — all exact, every time.
Built-in IP Protection
Every distributed .xagt file carries a unique cryptographic watermark. Unauthorized redistribution is instantly traceable. Your knowledge asset remains yours — no matter where it travels.
Cross-Platform by Design
One .xagt file runs identically on ChatGPT, Claude, Gemini, Llama, and every major AI platform. No rewrites, no vendor lock-in, no platform migration costs.
Open Standard
Like Linux for operating systems or CUDA for GPU computing, .xagt is an open standard for portable AI knowledge. Fully inspectable XML — no black boxes, no proprietary lock-in.
Physical AI Ready
The xagt runtime connects to IoT devices, smart home controllers, and robotic systems. Your knowledge agent can query real-world sensor data and trigger physical actions.
No-Code Builder for Experts
Lawyers, doctors, and consultants build production-grade AI without writing a single line of code. Upload documents, define your agent's persona, deploy to a global marketplace.
The .xagt File Skeleton
A declarative AI agent file composed of 9 semantic tags.
<xagt v="1.0" target="universal" lang="en">
<xdf>...</xdf> <!-- Tag Dictionary -->
<xhd>...</xhd> <!-- Header & Metadata -->
<xid>...</xid> <!-- Agent Identity -->
<xpl>...</xpl> <!-- Policy & Rules -->
<xkn>...</xkn> <!-- Core Knowledge -->
<xrt>...</xrt> <!-- Routing -->
<xwf>...</xwf> <!-- Workflows -->
<xcd>...</xcd> <!-- Executable Code -->
<xout>...</xout> <!-- Output Templates -->
</xagt>Every Part of the Agent, Precisely Defined
Each of the 9 tags serves a distinct structural role. LLMs parse the file like an API contract — the schema leaves nothing ambiguous.
Tag Dictionary
Declares all tag names and types upfront. Prevents LLM ambiguity by making the file's grammar self-describing.
Header & Metadata
File version, agent name, author, target platform, and runtime settings. The envelope of every .xagt file.
Agent Identity
The agent's persona, role, tone, and communication style. Defines how the AI presents itself and interacts with users.
Policy & Rules
Hard constraints, scope boundaries, and prohibited topics. Ensures the agent never strays beyond its defined purpose.
Core Knowledge
The main knowledge corpus. Structured Q&A pairs, procedures, domain data, and reference material the agent draws from.
Routing
Intent classification patterns that direct user queries to the right workflow. The agent's decision router.
Workflows
Multi-step conversation flows and guided processes. Defines exactly how the agent walks users through complex tasks.
Executable Code
Embedded Python or JavaScript for deterministic computations. The part of .xagt that never hallucinates.
Output Templates
Response format templates that ensure consistent, structured output. Define once — every answer follows the same professional format.
The Reading Order is the Design
LLMs retain the beginning and end of a file more strongly. .xagt strategically places knowledge to exploit this behavior.
Zone 1 — Primacy
The core knowledge area that the LLM reads first
Zone 2 — Middle
Classifies intent and routes to the appropriate workflow
Zone 3 — Recency
Execution instructions the LLM reads last and follows most strongly
3 Agent Execution Modes
Choose the engine type that best fits your use case.
Free-form conversational agent. Ideal for open-ended scenarios like author AI, consultation AI, or coaching AI.
Step-by-step guided agent. Optimal for ordered tasks like document drafting, contract review, and structured guidance.
Bulk processing agent. Used for generating analysis reports, data transformation, and large-scale document processing.
Who Builds with .xagt?
Any field where expertise is valuable and advice is expensive — .xagt turns that knowledge into a scalable, distributable AI.
Legal AI Assistants
Attorneys encode contract review playbooks, clause analysis logic, and jurisdiction-specific rules. Clients get instant, accurate legal guidance at a fraction of hourly rates.
Clinical Decision Support
Physicians encode diagnostic protocols, drug interaction checks via <xcd>, and triage flows. Nurses and patients get expert-level guidance at the point of care.
Financial Planning AI
CFOs and analysts encode investment criteria, tax optimization strategies, and portfolio rules. Clients interact with their advisor's logic — not just generic AI.
Expert Tutoring Agents
Teachers encode curriculum, problem-solving frameworks, and adaptive feedback logic. Students get personalized one-on-one instruction that scales infinitely.
Build xagt in 4 Simple Steps
From uploading sources to listing on the marketplace — anyone can start easily.
Upload Sources
Upload your knowledge documents (PDF, TXT, PPTX, etc.) to the Builder.
Build xagt
Choose from 4 strategies — pick the build that fits your scale and purpose.
Generate Deploy Link
Smart links auto-route between chatbot and web. Watermarks protect your IP.
List on Store
Register on the marketplace and start monetizing your knowledge.