Macchiato: Remote Control Messenger for Claude Code and Codex on iOS and Web

Macchiato launches a free AI messenger and open-source connector to remotely monitor and control Claude Code, Codex, OpenClaw, and Hermes from native iOS and we

tau · September 11, 2026

#Macchiato #ClaudeCode #Codex #OpenClaw #Hermes #AIAgent #CodingAgent #RemoteControl #DevTools

Macchiato: Remote Control Messenger for Claude Code and Codex on iOS and Web

On September 11, 2026, the Macchiato team (@macchiato_chat) officially announced Macchiato, a specialized AI messenger accompanied by an open-source connector designed to monitor and remotely control local AI coding and autonomous agents outside the desktop workstation. The platform enables developers to untether themselves from physical terminals while maintaining real-time communication with running agent processes across mobile devices and web browsers.

iPhone screen displaying Macchiato messenger app interface for remotely controlling Claude Code and Codex agents

Image source: Macchiato (@macchiato_chat) via X

While traditional coding agent workflows have predominantly confined developers to stationary command-line interface (CLI) sessions on local machines, Macchiato extends these operational feedback loops into an intuitive, chat-first messaging environment accessible anywhere.

Multi-Harness Compatibility and Open-Source Connector Architecture

The technical cornerstone of Macchiato is its native compatibility with leading AI coding and autonomous agent harnesses used across modern engineering workflows.

The system natively supports Anthropic's Claude Code, OpenAI's Codex, and autonomous open-source agent frameworks including OpenClaw and Hermes. Rather than forcing development teams to migrate into proprietary, vendor-locked cloud environments, Macchiato bridges directly into the agent setups and repository environments engineers already operate locally.

  • Open-source connector: Macchiato provides an open-source connector component that runs directly on the user's host machine or local workstation, facilitating an auditable communication bridge between local agent processes and the remote messaging layer.
  • Secure session pairing: Developers pair their active local sessions with the messenger service to receive real-time streams of agent execution logs, task milestones, and code modifications directly on their phones or browsers.
  • Tetherless supervision: Engineers can dispatch follow-up instructions, clarify project context, or query execution status on the go without configuring complex virtual private networks (VPNs) or heavy remote desktop tools.

Native iOS Client, Full Web Access, and Free Distribution Model

To accommodate varied remote working conditions, Macchiato provides cross-platform client interfaces across mobile and desktop browsers.

For Apple ecosystem users, a dedicated native iOS application titled 'Macchiato - Chat with Agents' is available on the Apple App Store, offering a fluid messaging user experience engineered specifically for mobile devices. In parallel, a full web application ensures consistent cross-platform access from any desktop or tablet browser.

  • Completely free service: Both the client applications (iOS and web) and the host connector software are released entirely free of charge with no subscription tiers or platform access fees.
  • Direct API token billing: While the Macchiato messaging service itself is free, the underlying large language model token consumption for active agents (such as Claude Code or Codex sessions) remains directly billed to the developer's individual provider API accounts.

Operational Prerequisites and Human-in-the-Loop Safety Governance

While remote agent management provides unprecedented operational flexibility, production deployments require careful attention to system prerequisites and execution safeguards.

To establish and maintain an active control session, the underlying agent processes (such as Claude Code, Codex, or Hermes) must be pre-launched and running on the developer's host machine or remote server, with the open-source connector properly paired before untethered operations begin.

  • Network connection reliability: Because instructions and telemetry flow across external messaging channels, maintaining stable host-to-client network connectivity is essential for continuous progress monitoring.
  • Human-in-the-loop policy management: Coding agents routinely exercise privileged capabilities, including file modifications and terminal shell command execution. Remote supervisors must establish explicit approval boundaries and safety policies to avoid unintended system actions when reviewing diffs away from primary desktop displays.

Sources