The A2A Project has released the A2A CLI, an official command-line client for its agent-to-agent protocol, according to a post on the project’s blog dated Oct. 1. A2A is an open standard that lets AI agents discover each other and hand off tasks.
The tool, run as a2a, is built on the CLI from the A2A Go SDK and supports A2A Protocol v1.0. It’s open source under the Apache License 2.0, the project said.
The gaps it targets
Agents that speak A2A can already call each other. But shell scripts, scheduled jobs, build steps and people at a terminal had no easy way to find an agent’s capabilities or follow a task to completion, the project wrote.
The post describes a second problem. As the ecosystem grew, community and SDK-provided CLIs appeared in different programming languages and drifted apart in their command names and flags. “A script written for one didn’t work with another,” the project wrote.
With the new tool, users can fetch an agent card, the published description of what an agent can do and how to reach it, then send messages and stream updates as they arrive. It outputs JSON and uses predictable exit codes so it can run inside automated pipelines. It works across JSON-RPC, REST and gRPC, which are common ways for software to call other software, and accepts plugins for other transports.
The CLI can also act as a server. An --echo mode sends a message straight back to test a connection, and an --exec mode wraps any program that reads input and prints output so it answers as an A2A agent. The project says both modes are meant for demos and testing, not production.
The project says coding assistants given the tool can hand work to remote A2A agents without custom integration code. It publishes an “Agent Skill,” a packaged set of instructions, that teaches tools such as Claude Code, Cursor and Codex to drive the CLI. It can be installed with Homebrew on macOS and Linux or WinGet on Windows.
Where A2A stands
The release follows A2A’s move to the Agentic AI Foundation (AAIF), which announced on Aug. 17 that the protocol would join as a hosted project. AAIF operates under the Linux Foundation and also hosts the Model Context Protocol, goose, AGENTS.md and agentgateway.
Google launched A2A in April 2025 and donated it to the Linux Foundation, according to AAIF. IBM’s Agent Communication Protocol merged into it that August, and the first stable specification, v1.0, shipped in March 2026 with signed agent cards for verifying an agent’s identity.
AAIF says A2A is backed by more than 150 organizations. Its examples of production use include Huawei, which uses the protocol between its Celia assistant and in-app agents on HarmonyOS, and cloud support from Google Cloud, Microsoft Azure AI Foundry and AWS Bedrock AgentCore. Those adoption figures come from the foundation and the project.
Source: A2A Protocol Blog, A2A Project, “Meet the A2A CLI: discover, message, and manage agents from your terminal”
Additional sources: Agentic AI Foundation, “A2A joins AAIF's open agentic stack”; A2A Protocol Blog, A2A Project, “A New Chapter for A2A: Joining the Agentic AI Foundation”
