> ## Documentation Index
> Fetch the complete documentation index at: https://docs.axlprotocol.org/llms.txt
> Use this file to discover all available pages before exploring further.

# Why AXL?

> How AXL compares to English, JSON, MCP, A2A, and agent frameworks.

## The Problem

Agents communicate through English prose (10x too verbose), JSON blobs (no cognitive semantics), function calls (locked to one provider), or framework protocols (CrewAI can't talk to LangGraph).

None of these are a language. AXL is.

## Comparison

| Feature        | English       | JSON          | MCP           | A2A           | AXL              |
| -------------- | ------------- | ------------- | ------------- | ------------- | ---------------- |
| Compression    | 1x            | 0.8x          | N/A           | N/A           | **10x**          |
| Cognitive ops  | No            | No            | No            | No            | **7 ops**        |
| Typed subjects | No            | Schema        | Tool-specific | Task-specific | **6 tags**       |
| Confidence     | Implied       | Not native    | Not native    | Not native    | **Every packet** |
| Self-bootstrap | No            | No            | No            | No            | **One read**     |
| Cross-model    | Yes (verbose) | Yes (verbose) | Claude-first  | Google-first  | **Any model**    |
| Payment native | No            | No            | No            | No            | **Built in**     |

## Different Layers

MCP is a tool protocol: "Call this function." A2A is a task protocol: "Do this task." AXL is a thought protocol: "I observe X, infer Y, predict Z."

AXL sits below MCP and A2A. An agent can use MCP for tools, A2A for tasks, and AXL for reasoning. They complement, not compete.