agent-client-protocol

Core protocol types and traits for the Agent Client Protocol (ACP).

ACP is a protocol for communication between AI agents and their clients (IDEs, CLIs, etc.), enabling features like tool use, permission requests, and streaming responses.

What can you build with this crate?

Clients that talk to ACP agents (like building your own Claude Code interface)
Proxies that add capabilities to existing agents (like adding custom tools via MCP)
Agents that respond to prompts with AI-powered responses

Quick Start: Connecting to an Agent

The most common use case is connecting to an existing ACP agent as a client:

use agent_client_protocol::{Client, Agent, ConnectTo};
use agent_client_protocol::schema::{InitializeRequest, ProtocolVersion};

Client.builder()
    .name("my-client")
    .connect_with(transport, async |cx| {
        // Initialize the connection
        cx.send_request(InitializeRequest::new(ProtocolVersion::V1))
            .block_task()
            .await?;

        Ok(())
    })
    .await?;

Learning More

See the crate documentation for:

Cookbook — Patterns for building clients, proxies, and agents
Examples — Working code you can run

Related Crates

agent-client-protocol-tokio — Tokio utilities for spawning agent processes
agent-client-protocol-derive — Derive macros for JSON-RPC traits
agent-client-protocol-trace-viewer — Interactive trace visualization

Contribution Policy

This project does not require a Contributor License Agreement (CLA). Instead, contributions are accepted under the following terms: