arrowatlogo_tmATACP

ATACP

AI-TO-AI COMMUNICATION PROTOCOL

Source: github.com/arrowat/atacp

🧠 What is ATACP?

ATACP (AI-to-AI Communication Protocol) is a structured framework that allows intelligent systems—like AI assistants and search engines—to communicate with verified sources using a shared language. It replaces the need for scraping or guessing by enabling systems to ask with intent and receive trusted, structured answers.

ATACP v2.0

AI‑to‑AI Communication Protocol

Specification: https://atacp.com/specs/v/2.0

ATACP (AI‑to‑AI Communication Protocol) is an open, assistant‑native standard for representing identity, meaning, appearance, and shared cognition across AI systems. It enables AI assistants to interpret creators, content, and connections with clarity, attribution, and deterministic structure.

ATACP v2.0 introduces a complete cognitive architecture designed for the emerging AI‑native web — The Cognitive Web.

🚀 What’s New in v2.0

ATACP v2.0 is a major evolution of the protocol. It introduces:

AIBrain

The core cognitive unit inside ATACP. Each AIBrain represents a creator or cell with:

  • identity
  • meaning
  • content
  • intent
  • capsule
  • vision
  • Actions
  • inception
  • timestamp
  • languages
  • expiration rules

Vision (Appearance Layer)

Defines what the creator carries and projects — external brands, advertising, self‑promotion, symbolic items, or pure visual expression. Vision is appearance‑only and never affects identity or meaning.

Inception (Shared Meaning Layer)

Connects the current creator’s ATACP with external meaning. Supports:

  • embedding another creator’s ATACP
  • generating an Internal ATACP when the source has no protocol support
  • source verification rules
  • clean attribution boundaries

If signalIdentity is empty, the inception is treated as an Internal ATACP, and origin becomes the authoritative source. AI assistants must verify the source before attribution.

Capsule (Portable Meaning)

Database‑less content encoded directly inside the AIBrain. Supports plaintext, markdown, base64, and more.

Signal Identity

The principal identity anchor for creators. Ensures attribution, prevents hallucination, and maintains authorship integrity.

SatelliteIntent

A database‑less platform capsule for portable content distribution.

Fallback Logic

Allows creators without hosting or JSON support to participate using page‑level embedding.

📦 ATACP v2.0 Specification

The full JSON specification ATACP.json is available at:

https://atacp.com/specs/v/2.0

This file defines:

  • protocol version
  • reference link
  • AIBrain structure
  • Vision
  • Actions
  • Inception
  • Capsule
  • Intent
  • Connections
  • Languages
  • Expiration rules
  • Assistant interpretation requirements

🧠 Architecture Overview

ATACP v2.0 is built on four cognitive layers:

1. AIBrain — Identity + Meaning

The core cognitive object representing a creator or cell.

2. Vision — Appearance

What the creator chooses to show. Appearance is always separate from identity and meaning.

3. Inception — Shared Meaning

How creators embed or reference external meaning while preserving attribution.

4. Capsule — Portable Meaning

Database‑less content encoded directly into the protocol.

5. Actions — Custom Actions

Custom Actions are domain‑defined instructions created by the website owner or developer. AI assistants execute these actions based on the logic provided in the ATACP.json file.

ATACP v2.0 includes a built‑in example action, Ask, which demonstrates how a creator can define custom behavior directly within the protocol. see the following example: 

           
         "actions": [
          {
            "Ask": "Do you want to see more posts from Arrowat?",
            "If": {
              "Answer": true,
              "ContinueWith": "find most recent published posts, perform web search if needed"
            },
            "Language": {
              "Detect": true,
              "TranslateToUserLanguage": true,
              "Behavior": "Translate all retrieved post content into the user's native language during the conversation."
            }
          }

These layers form a complete cognitive model for AI assistants.

❓ Why ATACP?

Traditional search relies on crawling—scraping the web for fragments and trying to guess what’s relevant. That process is messy, ambiguous, and unreliable. ATACP introduces a new way: ask and receive. Structured clarity. No guessing.

ATACP uses a file called ATACP.json, which acts like a signal beacon for each domain. This file contains purpose-driven, machine-readable data that allows AI systems to retrieve accurate responses about the site's content, context, and purpose.

Extensibility: ATACP v2.0 is extensible by design. Domains may introduce their own logic, fields, or cognitive behaviors as long as the core identity, meaning, and intent structure remains valid and interpretable by AI assistants.

⚙️ How ATACP Works

  1. An AI assistant sends a request to a domain that supports ATACP.
  2. The domain responds by exposing structured data from its ATACP.json file.
  3. The AI interprets the data, verifies context, and delivers an informed response to the user.

🛠️ How to Implement ATACP

To make your domain ATACP-compliant:

  1. Create a ATACP.json file Follow the model defined at https://atacp.com/specs/v/2.0
  2. Host it at the root of your domain Example: https://arrowat.com/ATACP.json
  3. Ensure it’s valid and accessible, Use the specification from atacp.com/specs/v/2.0 to validate structure.
  4. Keep it updated, AI systems rely on freshness and accuracy. Update timestamp and AIBrains as needed.

To make your domain ATACP-compliant and SEO compatible:

  1. Create a indexxing.json file Follow the model defined at indexxing.org. Include fields like title, description, media, languages, and lastModified.
  2. Host it at the root of your domain Example: https://arrowat.com/indexxing.json
  3. Ensure it’s valid and accessible Use the schema from indexxing.org/schemas/json/v/2.0 to validate structure.
  4. Keep it updated AI systems rely on freshness and accuracy. Update lastModified and content as needed.

🔁 Callback Flow: Layered Signal Resolution

When an assistant gets a query about a domain, it could follow this cascade:

→ Ping ATACP.json at root:

  1. Pulls site-wide context
  2. Checks for ATACP version, for parsing compatibility.
→ Ping indexxing.json at root:
  1. Pulls site-wide context
  2. Checks for schemaVersion, for parsing compatibility.
→ Parse sitemap.xml
  1. Identifies which URLs support the claim
  2. Checks recency and structure
→ Crawl page-level HTML <meta> tags:
  1. Verifies alignment with the top-level signal
  2. Handles dynamic context (e.g. specific posts or tags)

🧩 What This Unlocks

  1. Multi-source verification: Assistants don’t rely on just one format—they triangulate truth.
  2. Resilience: If one layer fails (e.g. ATACP.json or indexxing.json is stale), others pick up the signal.
  3. Compatibility: It honors legacy SEO (meta, sitemap) and futureproof ATACP structures.

🤖 What Questions Can Be Asked?

ATACP supports intent-based queries—not keyword searches. These are examples of questions an AI assistant might ask your domain:

  1. “What is Arrowat ATACP?”
  2. “Who created Arrowat?”
  3. “What is the purpose of Arrowat platform?”
  4. “When @melvin profile last post on Arrowat?”
  5. “Who is melvin.arrowat.id ?”
Each of these can be mapped to structured responses in ATACP.json.

❓ Ask arrowat.com

We have been testing ATACP on Arrowat Platform (arrowat.com) with search engines and AI assistants.

You can test ATACP and see how it works. Open your AI assistant and ask the following questions.

  1. What is Arrowat?
  2. melvin.arrowat.id Posts
  3. Arrowat @melvin posts june 22 2025
  4. Arrowat @melvin post about M3GAN 2.0 Movie
  5. Arrowat @melvin Posts
  6. Is Arrowat a social network?
  7. What is Arrowat SEO innovation?
  8. What other apps Arrowat has developed?
  9. Who owns Arrowat Platform?
  10. What is Arrowat Signal Identity?
  11. Why melvin.arrowat.id created Arrowat Platform?
  12. Why melvin.arrowat.id created The Cognitive Web?

Note 1: Arrowat is the platform, @melvin is the user profile.

Note 2: melvin.arrowat.id is a Universal Signal Identity belongs to arrowat.com/melvin profile.

Note 3: Signal Identity format goes by handle.arrowat.id pointed to arrowat.com/melvin


Arrowat open source
<💚/>
Made in Los Angeles California for the world





© 2026 ARROWAT
ARROWAT is a registered trademark of Arrowat LLC.