acp

New

ACP is an open protocol and toolkit designed to enable communication and coordination between AI agents, applications, and humans. The repository provides the protocol specification, server and client SDKs, example agents, and documentation so developers can build interoperable, multimodal agents that discover each other, send rich message parts (text, files, media, code), and collaborate on tasks. It is intended for teams building agents or agent platforms rather than end users. The codebase includes an OpenAPI specification, a Python server implementation with client libraries and model definitions, a TypeScript client SDK, ready-to-run examples such as a retrieval-augmented LlamaIndex agent, and guides including a hands-on short course. The project supports streaming and background responses, session state, run lifecycle control, and extensions for production deployment.

Stars

652

App URL

https://github.com/i-am-bee/acp

Github Repository

https://github.com/i-am-bee/acp/blob/main/README.md

Features

ACP defines structured multimodal messages composed of Message and MessagePart primitives and adds metadata types such as TrajectoryMetadata and CitationMetadata for tracking reasoning and sources. It specifies agent manifests for capability discovery and a Run lifecycle that supports synchronous, streaming, and background execution. The protocol includes an Await mechanism so agents can pause for external input, distributed sessions to maintain continuity across server instances, and first-class session state handling. Toolkit components include an OpenAPI specification, a Python SDK with a server implementation and client libraries, a full TypeScript client SDK, documentation and examples, a RAG LlamaIndex example, and guidance for high availability deployments using centralized storage. The repo also documents message role parameters and supports file and media parts.

Use Cases

ACP reduces fragmentation by providing a common protocol that lets agents developed in different frameworks interoperate, discover capabilities, and compose complex workflows. Developers can use the OpenAPI spec and SDKs to quickly implement agents and clients, follow the quickstart to run an example echo agent, and examine examples for retrieval-augmented generation. The Await and session features let agents pause for input and maintain conversation history across runs, which simplifies building stateful and interactive agents. Metadata features improve traceability of multi-step reasoning and source attribution. High availability and distributed session guidance enable scalable production deployments, and the protocol is used by the BeeAI platform to run and share agents.

acp

Basic Information

Links

Categorization

App Details

Categories

Similar Listings

virtual lab

mnemo

EdgeChains

RAG Agents Accelerator

LLM Zero to Hundred

Featured Listings

Terry Bison Ranch

The Singapore Flyer

Tags

More Filters

acp