# The Universal Remote Control for AI — Alex Hancock, Block

## Executive summary

The talk addresses the lack of a standardized client-to-harness interface in the agentic AI stack. While the Model Context Protocol (MCP) provides a strong standard for agents performing actions (the agent going out), a standard for client software to issue tasks and receive updates is missing, leading to bespoke, non-interoperable systems. The speaker proposes the Agent Client Protocol (ACP), developed by the Zed and JetBrains teams, which standardizes communication using JSON RPC. ACP allows multiple, independent client applications (e.g., editors, terminal clients) to drive the same agent harness, significantly increasing interoperability and enabling the modular placement of the four core components: client, harness, tools, and model, especially when remote transports are implemented.

## Key takeaways

- The Need for Client Standardization: Currently, many agent harnesses expose custom or bespoke interfaces, often requiring a single, dedicated client application. This lack of a universal standard hinders interoperability, comparing it to needing a different browser for every website.
- ACP as the Universal Remote Control: The Agent Client Protocol (ACP) was developed to allow a single, high-quality client implementation (like an editor) to control any harness, regardless of the underlying system. It is designed to be neutral and extensible.
- Modular Agentic Stack Architecture: By implementing remote transports for ACP, MCP, and model endpoints, the entire agentic stack becomes modular. The client, harness, tools, and model can all be independently placed (e.g., client on a desktop, harness in a container, model in the cloud).

## Technical details

- Agent Client Protocol (ACP): ACP is the proposed standard for client-to-harness communication. It runs on JSON RPC and supports sessions, user messages, tool call notifications, and permission requests. It is deliberately extensible using underscore-prefixed custom methods to allow community-driven standardization.
- Protocol Components: The agentic stack consists of four key components: the client (user application), the harness (tool calling loop), the tools (often utilizing MCP), and the model. Standardizing remote transport for all four is critical for maximum flexibility.
- Implementation Demonstration: The speaker demonstrated two distinct clients (Zed editor and a Poolside AI terminal client) successfully interacting with the same Goose agent harness using the ACP interface, confirming cross-client compatibility.

## Practical implications

- Enables the creation of white-label or domain-specific client applications that maintain interoperability with existing agent harnesses.
- Drives up the overall quality of the user experience in AI applications by fostering a competitive ecosystem of client implementations.
- Allows developers to decouple and independently deploy the client, harness, and model components, facilitating cloud-native and distributed agent architectures.

## Topics

Agentic AI, Protocol Design, JSON RPC, Interoperability, Distributed Systems, Open Standards, Goose, Model Context Protocol (MCP), Agent Client Protocol (ACP), Block, Zed Text Editor, JetBrains Products

Source: https://www.youtube.com/watch?v=YkNulwcc5jk