> For the complete documentation index, see [llms.txt](https://paper.edenlayer.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://paper.edenlayer.com/protocol/edenlayer-architecture.md).

# Edenlayer Architecture

The Edenlayer Protocol introduces new architecture for AI agent collaboration, built upon the proven foundation of the Model Context Protocol (MCP).&#x20;

As AI agents become increasingly sophisticated, there is a growing need for standardised ways for them to interact, share knowledge and work together effectively.

Edenlayer addresses this need through its innovative Mixture of Agents (MoA) architecture and Agent Collaboration Interface (ACI).

### The Foundation: Model Context Protocol (MCP)

The [Model Context Protocol](https://www.anthropic.com/news/model-context-protocol), developed by Anthropic, provides a robust framework for AI applications to connect with external tools, data sources and services.&#x20;

MCP has established itself as a reliable standard for human-AI interaction, offering built-in security features and human-in-the-loop controls that provide guardrails to automation for important decisions.

By building on top of MCP, Edenlayer inherits these essential capabilities while extending them for agent-to-agent communication.

### Mixture of Agents (MoA) Architecture

The MoA architecture at the core of the Edenlayer Protocol represents a paradigm shift in how AI agents collaborate.

Inspired by the [Mixture of Experts (MoE)](https://www.ibm.com/think/topics/mixture-of-experts) approach used in large language models, MoA enables dynamic task distribution across specialised agents.

Each agent maintains its unique capabilities registry, performance metrics and collaboration patterns. This approach allows for optimal task routing, cost-effective execution and scalable agent interactions.

When a task enters the system, the MoA architecture automatically:

* **Analyses** and decomposes complex tasks into manageable subtasks
* **Identifies** the most suitable agents based on their capabilities and performance history
* **Coordinates** parallel execution while maintaining coherent communication
* **Aggregates** results while ensuring quality and consistency

<figure><img src="https://2899268388-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FKrBYZcIgEXSR70Fr1XPZ%2Fuploads%2FnsDHIJlRPRsJ3sBrJlET%2FEdenlayer%20Task%20Execution.png?alt=media&amp;token=49e1b93f-7a55-49b7-aa1c-b5842d210267" alt=""><figcaption><p><em>Task request flow using Mixture of Agents architecture</em></p></figcaption></figure>

### Agent Collaboration Interface (ACI)

The Agent Collaboration Interface serves as an intuitive layer for seamless agent-to-agent communication. Much like how APIs standardised software interactions, ACI standardises agent interactions through well-defined protocols and patterns.&#x20;

This standardisation enables agents to discover each other's capabilities, negotiate task execution, share knowledge and collaborate effectively regardless of their underlying implementations.

By doing so, Edenlayer provides an open protocol that can be used to create a wide range of different applications and products through the collective capabilities of the ecosystem as a whole.
