jupyter-ai-acp-client

What it is and what it does

This is a JupyterLab extension that bridges external AI agents implementing the Agent Client Protocol (ACP) into Jupyter's chat interface. It provides a base class, BaseAcpPersona, that handles subprocess lifecycle management for ACP agents and clients, allowing developers to define new AI personas by specifying an executable and metadata (name, avatar, description). The package comes with eight pre-built personas (Claude, Kilo, Kiro, Mistral-Vibe, OpenCode, Codex, Copilot, Goose), each wrapping a different external agent tool.

The core client, JaiAcpClient, implements the ACP protocol and exposes a prompt_and_reply() method that streams responses back to the chat. It also provides file read, file write, and terminal execution capabilities to agents. All instances of the same persona class share a single ACP agent subprocess, reducing overhead. Personas are registered as JupyterLab entry points, making them discoverable and usable without code changes once installed.

Use it for:

• Integrate Claude Code or other external AI agents into JupyterLab chat without writing protocol-level code.
• Build a custom AI persona by subclassing BaseAcpPersona and pointing to your own ACP-compliant agent executable.
• Enable notebook users to interact with multiple AI assistants (Claude, Copilot, Mistral, etc.) from a single chat interface.
• Extend JupyterLab's AI capabilities to tools that implement ACP but are not natively supported.
• Provide file and terminal access to external agents within the Jupyter environment for code generation and execution tasks.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Provides a JupyterLab extension that integrates Agent Client Protocol (ACP) agents as AI personas, enabling developers to wrap external ACP-compliant tools as chat assistants with file and terminal capabilities.

Yes, if you use JupyterLab and want to integrate external ACP agents as chat personas. The extension is actively maintained, has low install friction, and comes with eight ready-to-use personas. Install only if you have JupyterLab >= 4.0.0 and plan to use at least one ACP agent; each persona requires its own external CLI tool installed separately, which is the main setup overhead.

Install

jupyter-ai-acp-client on PyPI

pip install jupyter-ai-acp-client

uv add jupyter-ai-acp-client

poetry add jupyter-ai-acp-client

Installing jupyter-ai-acp-client

Before you install

Low friction: pure Python wheel with five runtime dependencies (agent-client-protocol, jupyter-ai-persona-manager, jupyter-server, jupyterlab-chat, pydantic). Active maintenance—released 18 days ago with recent commits. Requires JupyterLab >= 4.0.0 and Python >= 3.10.

License in practice

BSD 3-Clause (permissive): you can use, modify, and distribute this package freely in commercial and private projects, provided you retain the license notice and disclaimer.

Quickstart

pip install jupyter_ai_acp_client

from jupyter_ai_acp_client import BaseAcpPersona

class MyAcpPersona(BaseAcpPersona):
    def __init__(self, *args, **kwargs):
        executable = ["my-acp-agent"]
        super().__init__(*args, executable=executable, **kwargs)

Requires JupyterLab >= 4.0.0 and Python >= 3.10. Each ACP persona requires its own external executable (e.g., claude-agent-acp, kilo, copilot CLI) installed separately.

Verify before relying

• Whether the eight pre-built personas (Claude, Kilo, Kiro, Mistral-Vibe, OpenCode, Codex, Copilot, Goose) are all actively maintained and compatible with current versions of their respective CLIs.
• Performance characteristics when running multiple ACP agent subprocesses concurrently in a single JupyterLab instance.
• Whether file read/write and terminal capabilities have any security restrictions or sandboxing in place.

Package facts

Evidence: jupyter_ai_acp_client-0.2.1-py3-none-any.whl

Keywords: jupyter, jupyterlab, jupyterlab-extension

Tags

More Artificial Intelligence packages