OpenClaw Is Now Available on Android and iOS

OpenClaw, the free and open-source AI agent platform that became one of the fastest-growing projects on GitHub earlier this year, shipped its first official native mobile apps on June 29, 2026. Both an iOS and an Android version are now available through the Apple App Store and Google Play Store respectively, giving the project's growing user base a way to interact with their self-hosted AI agents from anywhere, without being tethered to the desktop environment where OpenClaw has lived until now.

To understand what this launch actually means, it helps to understand what OpenClaw is and how it works, because the mobile apps do something genuinely different from most AI assistant apps. They are not standalone chatbots that send your requests to a company's cloud server. They are companion nodes that connect to a self-hosted OpenClaw Gateway you run on your own hardware, turning your phone into a secure remote control for the AI agent network you have already built at home or in your lab. That distinction matters a lot, and it is the specific reason Apple approved the app when it has historically been resistant to agentic AI tools in the App Store.

What OpenClaw Is and Why It Went Viral

OpenClaw originally started life under a different name. The project began as Clawdbot, briefly became Moltbot, and eventually settled on its current branding as it grew into one of the more distinctive entries in the personal AI agent space. OpenClaw is a personal AI assistant that users run on their own devices, with an architecture built around the principle that users should retain full control over their gateway, API keys, configuration, and permissions at all times.

The project was created by Austrian developer Peter Steinberger, who built OpenClaw as an open-source framework for running AI agents on your own hardware. It exploded into broader public awareness in January 2026 when it picked up over 60,000 GitHub stars in 72 hours, a rate of organic adoption that few open-source projects achieve. The project has since grown to cross 1.17 million weekly npm downloads, making it a genuine phenomenon in the developer community rather than a niche tool with a small following.

Part of what drove that initial viral moment was a memorable stunt. OpenClaw went viral around the launch of MoltBook, a social media site purportedly populated entirely by AI agents. The spectacle later turned out to have been partially the work of humans impersonating agents, though the underlying demonstration of what agent-to-agent interaction could look like captured enough imagination to drive genuine downloads and deployments regardless of how the theatrical scaffolding was constructed. In February 2026, Steinberger announced he had joined OpenAI, a move that gave the project a significant credibility boost and secured unspecified support from OpenAI for the OpenClaw Foundation that continues to steward the project.

"Agents in your pocket. Run agents from wherever your thumbs are."
- OpenClaw official launch announcement, June 29, 2026

How the Mobile Apps Actually Work

The most important thing to understand about the OpenClaw iOS and Android apps is what they are not. They are not standalone AI assistants. They do not send your messages to OpenClaw's servers, process them in the cloud, and return a response. The app itself does very little without a user-controlled gateway already running on separate hardware. This companion-node architecture is both the defining technical characteristic of the apps and the specific reason Apple approved them when it has previously blocked agentic AI tools from the App Store over concerns about autonomous system access.

The architecture works as follows. Users run an OpenClaw Gateway on a Mac, Linux machine, or Windows PC. That gateway is the actual control plane for their AI agent network, the component that connects user requests to AI models, tools, skills, and workflows. The iOS or Android app then connects to that gateway over WebSocket, pairing via QR code or a setup code in a process that takes a few minutes once the gateway is already running. Once connected, the phone functions as a secure node within the user's personal agent infrastructure, able to send requests, receive responses, approve agent actions, and monitor workflow status from wherever the user happens to be.

Because the app is purely a WebSocket client connecting to a gateway the user controls on their own hardware, Apple is not approving an autonomous AI agent with broad system access. It is approving a remote control application. That framing is what allowed OpenClaw to clear the App Store review process, which has been notably resistant to agentic AI applications for years. The App Store listing for the iOS version explicitly states "Data Not Collected," a designation that reflects the self-hosted architecture's fundamental property: when your gateway runs on your own machine, there is no third-party cloud service collecting data from it.

What You Can Actually Do With the Apps

OpenClaw described the core functionality of the mobile apps in its launch announcement with three concise framing points: native mobile apps, finally; agents in your pocket; channels, tasks, replies on the go. The specifics behind each of those points are worth unpacking individually.

Chat and Real-Time Voice

The most straightforward capability is text chat: users can interact with their OpenClaw assistant directly from the mobile interface using the same conversational input they would use on desktop. Beyond text, both apps support a Talk mode that enables real-time voice conversations with the agent, including a background Talk mode that can continue processing while the app is not in the foreground. For users who have built agents that run extended workflows, the background mode means they do not need to keep the app open on screen to receive status updates or voice responses as the agent works through tasks.

Remote Action Approvals

Every action an agent wants to take on the gateway requires explicit user approval before it proceeds. The mobile apps surface these approval requests as push notifications, allowing users to review, approve, or deny agent actions from their phone even when they are nowhere near the machine running the gateway. This human-in-the-loop approval mechanism is central to OpenClaw's philosophy and addresses one of the more common concerns about autonomous AI agents operating without oversight. The agent can propose actions, but it cannot execute consequential ones without an explicit approval from the user on whatever device they happen to be using.

Selective Device Capability Access

Both apps allow users to selectively grant the OpenClaw agent access to mobile device capabilities that are not available to a desktop gateway: the camera, screen, location, photos, contacts, calendar, and reminders. Each of these access grants is managed through the operating system's native permissions model rather than through OpenClaw's own permission system, which means users interact with the standard iOS or Android permission dialogs they are already familiar with, and can revoke access at any time through standard system settings. No capability is enabled by default; everything requires an explicit user opt-in.

The iOS app additionally supports sharing content directly from other apps into OpenClaw, making it possible to pass a link, image, or piece of text from any iOS app into the agent's context without having to manually copy and re-enter it. The Android version supports launching OpenClaw via the Google Assistant trigger, so holding the home button or saying a specific voice command routes directly into the OpenClaw app.

Push Notifications and Workflow Status

Push notifications keep users updated on the status of ongoing workflows and connected gateway nodes even when the app is in the background. For users who have set up extended automation workflows that run for minutes or hours, this persistent status visibility removes the requirement to stay at the desktop to know what the agent is doing and when it needs attention.

How iOS and Android Differ

The Android App's Rough Launch

Coverage of the launch has been honest about an important caveat: the Android version is in considerably rougher shape than the iOS release at the time of writing, and the early user response reflects that gap clearly. The Android app received a 2.2 star rating in the Play Store at launch, with early reviews including descriptions of the app as buggy, unusable, and reports of consistent pairing failures preventing users from connecting to their gateway at all. One review went as far as calling it the worst app I've ever used in my entire life, which is harsh language that nonetheless reflects a pattern of real functional problems rather than mere aesthetic preference.

The visual design drew its own criticism separately from the functional issues. The app's preview screenshots show an extremely bare-bones design that literally overlaps with the status bar, placing the OpenClaw header behind the clock and notification icons. One reviewer who installed the app on a Galaxy Z Fold 7 noted it looked better in actual use than in the screenshots, though still fell short of the polish level most Android users expect from a public Play Store release in 2026.

It is worth contextualizing this against OpenClaw's historical approach to software design. OpenClaw has never put form over function. The project is, at its core, a command-line interface for most of its desktop existence on Windows and macOS. The iOS app looks and functions more polished than the Android release, suggesting the team may have prioritized the Apple platform for the initial launch, with Android following in a state that is genuinely functional for users who can get past the pairing issues but not yet production-ready by typical app quality standards.

The foundation has been transparent that iteration is the plan here, and the project's open-source nature means community members can and do contribute fixes. Given the pace at which OpenClaw has historically shipped, it would be reasonable to expect meaningful Android improvements within weeks of this initial release rather than months.

Why Apple Approved an Agentic AI App

The App Store approval for OpenClaw's iOS app deserves specific attention, because it has been explicitly framed by analysts as the first agentic AI tool to clear Apple's historically resistant App Store review process. Apple has previously blocked numerous agentic applications over concerns about autonomous AI access to device capabilities and system functions, particularly following a period of heightened scrutiny around vibe-coded applications with poorly defined permission models.

The reason OpenClaw cleared that bar comes down to architecture. Because the app is purely a remote control for a gateway you run on your own hardware, Apple is not approving an autonomous AI agent with system access. It is approving a WebSocket client. From the App Store review process's perspective, the OpenClaw app does not actually do anything independently. Without a user-controlled gateway already running, it is an inert connection interface with no autonomous capability to collect data, execute system actions, or reach out to third-party services. Every capability it can access, whether camera, location, or contacts, requires the same user permission grants that any other iOS app must obtain through the standard system permissions dialog.

This architecture is the specific answer to a question that has stumped other agentic AI developers trying to get iOS apps approved: how do you put a capable AI agent on a phone in a way that Apple's review team will accept? OpenClaw's answer is to not put the agent on the phone at all. The phone is the interface; the agent lives somewhere the user already controls.

The Local-First Philosophy and What It Means for Privacy

OpenClaw's defining characteristic has always been its local-first architecture, and the mobile apps extend that philosophy to a new form factor rather than compromising it in the interest of a simpler cloud-based setup. OpenClaw's defining philosophy is what the project calls a local-first strategy: users retain full control over their gateway, API keys, configuration, and permissions. Device capabilities are gated entirely by the operating system's own permissions model, and can be enabled only for the features a user actively chooses to use. This architecture means the mobile app itself does not independently store data or relay it through OpenClaw's servers. The Gateway remains the authoritative control plane; the phone is simply a new kind of node attached to it.

The privacy implications of this design are meaningful and distinguishable from the approach most commercial AI assistant apps take. When a user chats with a mainstream AI assistant on their phone, that message typically travels to a company's cloud infrastructure, is processed by models running on that company's servers, and the response comes back through the same channel. The company has access to the query and, in many configurations, may use it to improve models or for other purposes specified in their terms of service.

With OpenClaw's mobile implementation, the user's message travels from the phone to the gateway on their own hardware. The gateway processes the request using whatever AI model the user has configured, with whatever API keys they have supplied, and returns the response through the same private channel. If the gateway is running locally at home on a network with no external exposure, the entire transaction remains within the user's own infrastructure. The App Store's Data Not Collected designation for the iOS version accurately describes this reality: there is no data for OpenClaw to collect because the data never passes through OpenClaw's systems.

How to Get Started With the Mobile Apps

Using the OpenClaw mobile apps requires an existing OpenClaw Gateway installation, which is the necessary prerequisite before the app can do anything useful. For users who have not set up OpenClaw before, the onboarding process involves running openclaw onboard in a terminal, which walks through gateway setup, workspace configuration, and channel connections step by step. Once the gateway is running on a Mac, Linux, or Windows machine, connecting the mobile app is a matter of opening the iOS or Android app and scanning the QR code displayed by the gateway, or entering a setup code manually.

• Install the OpenClaw Gateway on a Mac, Linux, or Windows machine by running openclaw onboard in a terminal
• Download the OpenClaw app from the App Store on iOS (requires iOS 18 or later) or the Play Store on Android
• Pair the mobile app to the gateway by scanning a QR code or entering the displayed setup code
• Selectively enable device capabilities including camera, location, contacts, and calendar through standard OS permission prompts
• Chat, use Talk mode for voice interaction, or approve gateway actions directly from the mobile interface

For existing OpenClaw users who have already built workflows and configured channels, the mobile app functions as an immediately available companion without any additional gateway configuration. The gateway does not need to be reconfigured to support the mobile connection; it handles the new node as it would any other connection to the OpenClaw network.

Where OpenClaw Sits in the Broader AI Agent Market

OpenClaw's mobile launch happens at a moment when agentic AI has become one of the most active categories in the entire AI product market. Agents are now embedded across the AI landscape and are showing up in more places by the day. The mobile deployment of AI agents, specifically giving users the ability to manage and interact with agents from their phones without sacrificing privacy or control, is a genuinely underserved area that most existing agent platforms have not addressed in a way that satisfies privacy-conscious users.

The competitive interest in OpenClaw's approach has already become visible at a company level. Google is reportedly building its own 24/7 personal agent to compete with OpenClaw directly, a signal of how seriously the search giant is taking the local-first, persistent personal AI agent category that OpenClaw essentially pioneered. OpenAI's own interest in the project, formalized through Peter Steinberger's hiring and the subsequent support for the OpenClaw Foundation, represents a different kind of competitive engagement, one in which the largest AI company in the world is aligned with an open-source project it did not create but now has commercial reasons to support.

The comparison to how OpenAI handled its own agentic mobile integration is instructive. The mobile apps function as companion nodes that pair with a user's existing OpenClaw installation, much like the Codex controller OpenAI has embedded into the ChatGPT app. The parallel framing suggests the companion-node pattern, where a phone app serves as a remote control rather than a standalone AI endpoint, may be emerging as the dominant design pattern for mobile AI agent interfaces, at least in the near term before regulatory and technical environments around fully autonomous mobile AI agents become clearer.

What Users Have Built and What They Are Doing With It

OpenClaw's open-source community has been building with the platform for most of the year, and the range of use cases that have emerged speaks to the flexibility of an agent framework that users configure themselves rather than one with a pre-defined set of supported workflows. OpenClaw users have put it to work in everything from coding to meal planning, though some have reported less-than-desirable results in more complex workflow configurations.

The mobile launch extends all of those existing use cases to a new form factor. An agent configured to help manage a development environment can now be interacted with from a phone while away from the desk. An agent configured to help with research workflows can receive a link shared directly from the iOS Share Sheet and begin processing it while the user is on the move. An agent running scheduled monitoring tasks on the gateway can push notification updates to the phone in real time without requiring any manual check-in.

For users who push the platform harder, the combination of phone-level device access, camera, location, contacts, calendar, and reminders, with the agent capabilities running on the gateway, opens up workflows that were not possible when agents could only operate within the desktop environment. An agent with access to a user's calendar and location could theoretically manage scheduling decisions contextually based on where the user actually is, a capability class that becomes practical only when the phone is a first-class node in the agent network rather than an afterthought.

What to Watch Going Forward

The most immediate thing worth watching is whether the Android app improves quickly enough to retain the users who downloaded it on launch day, had a poor experience, and left negative reviews. A 2.2 star rating at launch is a rough start, and the open-source community's ability to contribute fixes in parallel with the OpenClaw Foundation's own engineering work will determine how quickly the Android experience reaches the standard the iOS version already partially achieves. Early issues around the gateway pairing flow appear to be the single highest-priority problem to resolve, since the pairing step is the prerequisite for any other functionality.

Beyond the immediate quality issues, the broader question is whether OpenClaw's local-first, privacy-respecting approach to mobile AI agents attracts users who have been waiting for exactly this architecture, or whether the gateway setup requirement creates enough friction that most users default to mainstream cloud-based AI assistants that trade privacy for a simpler onboarding experience. The viral growth earlier this year suggests genuine appetite for OpenClaw's philosophy exists in the developer and privacy-focused community. Whether that appeal extends to a broader mainstream audience that encounters the gateway setup requirement before it can even start chatting is the question the mobile launch will begin to answer over the coming months.

Related Topics: #OpenClaw #AIAgents #OpenSource #iOS #Android #OpenAI #PrivacyAI #LocalFirstAI #Technology #ArtificialIntelligence