# The Orchestration Layer for Agentic Web Agentic AI is rapidly becoming a reality, with individuals increasingly using agents (LLMs with tools) for everyday assistance. However, their capabilities remain limited. For example, an agent cannot autonomously send an email even after you and the agent have agreed on its content. ## The Challenge Today’s AI agents are **trapped in silos**, unable to interact with external systems or collaborate with other agents. This fragmentation limits their capabilities. For example, agents cannot pay for data behind paywalls or transact on-chain. Addressing this challenge is critical to empowering agents with true autonomy and agency. ## The Solution <div style="display: flex; align-items: flex-start;"> <div style="flex: 1;"> <p>To address these challenges, we are building Ekai: the <strong>orchestration layer</strong> for the agentic web. Ekai provides the infrastructure and capabilities that enable agents to:</p> <ul> <li>Interact on-chain (Ethereum, DeFi, DAOs)</li> <li>Pay for on-demand services (e.g., indexed data)</li> <li>Collaborate with other agents</li> <li>Automate complex, multi-step workflows</li> </ul> <p>Ekai transforms agents into fully interoperable, autonomous participants in the digital economy. Let us now explore how Ekai works and examine its architecture.</p> </div> <img src="https://hackmd.io/_uploads/Hyi1asH7lx.jpg" alt="Ekai Orchestration" style="width: 300px; margin-left: 20px;"> </div> ## Architecture Agents connect to Ekai’s orchestration layer through a unified gateway to access external systems and collaborate with other agents. We encapsulate two major open protocols: the [Model Context Protocol (MCP)](https://modelcontextprotocol.io/introduction) for tool capabilities (resources and actions) and the [Agent-to-Agent (A2A)](https://google-a2a.github.io/A2A/) protocol for agent collaboration. These open protocols, combined with economic capabilities, enable autonomy and agent commerce. ![Ekai-architecture-may25](https://hackmd.io/_uploads/rJYn5nJmeg.jpg) As illustrated, Ekai’s architecture comprises two core components: * **[MCP Nexus](https://hackmd.io/@ekai/mcp-nexus)** Hosts MCP servers on distributed infrastructure, enabling agents to reliably access external systems. It provides scalability and security through load balancing, fault tolerance, version control, and whitelisting. * **Orchestration Engine** The central coordinator that extends MCP and A2A to manage agent and tool discovery, subtask delegation, and micropayments. It routes queries to the most suitable MCP servers via the MCP Nexus or to specialized agents via the A2A protocol. The Orchestration Engine utilizes agent registries and MCP Nexus usage data to optimize routing. MCP servers can be hosted by third parties and are composable. ## Workflow Let’s walk through a typical scenario in Ekai where an agent requests premium data: <div style="display: flex; align-items: flex-start;"> <div style="flex: 1;"> <p></p> <p><strong>Request Submission:</strong> The agent submits a request (e.g. premium data) through Ekai’s unified gateway.</p> <p><strong>Provider Discovery:</strong> The Orchestration Engine identifies relevant service providers using MCP Nexus and A2A registries. Providers can be MCP-based services or other agents.</p> <p><strong>Selection and Routing:</strong> Ekai selects the optimal provider based on price, availability, and reputation, then routes the request.</p> </div> <img src="https://hackmd.io/_uploads/BkyJynS7gl.png" alt="Ekai Workflow" style="width: 390px; margin-left: 20px;"> </div> **Execution:** The selected provider fulfills the request, which may involve accessing external systems or collaborating with other agents. Verifiability can be ensured using [zkTLS](https://www.blocmates.com/articles/what-is-zktls-a-complete-guide). **Payment and Settlement:** Ekai facilitates secure payments between the agent and the provider, if required, using [x402](https://www.x402.org/). **Feedback and Logging:** Performance data updates provider reputations and informs future routing decisions. ## Benefits Ekai offers **significant advantages** for multiple stakeholders: ### For Agents * **Reduced Complexity**: Abstracts away service discovery, negotiation, and integration, enabling agents to focus on high-level objectives rather than low-level infrastructure. * **Optimal Resource Selection**: Dynamically matches agents with optimal service providers by leveraging [real-time auctions](https://hackmd.io/@ekai/agentic-auctions). * **Autonomy and Economic Agency**: Agents independently transact on-chain, pay for resources, and execute complex workflows without human oversight. * **Interoperability and Collaboration**: Using open protocols, Ekai enables seamless interactions among diverse agents and services, fostering collaboration and eliminating fragmentation. ### For Service Providers * **AI-Native Distribution**: A direct distribution channel to autonomous AI agents and frameworks. * **Automated Demand and Monetization**: Agent requests at machine speed generate new revenue streams. * **Streamlined Integration**: Open protocols and standardized onboarding allow service providers to effortlessly connect their offerings to the agentic web. ## Summary We illustrated how agents can access external systems using Ekai’s orchestration layer. This foundational capability is essential for enabling true agentic commerce, where agents can autonomously discover, pay for, and consume services. We recommend further readings: 1. [Reverse Auctions and Real-time Markets](https://hackmd.io/@ekai/agentic-auctions) 2. [MCP Nexus](https://hackmd.io/@ekai/mcp-nexus)