Loading...

The first agentic mobile app with an on device Companion to manage the app and your agent teams.

The Companion also acts as your personal Chief of Staff.

On-device agents possess full agentic capabilities: advanced tools, skill support, persistent memory, and recurring schedules—all with local file access and external connectivity via MCP. These personal agents are supercharged by native mobile superpowers, giving them permissioned access to your camera, location, and contacts others.

Access your remote coding agents, OpenClaw, and complex agentic workflows through a single, central communication hub.

Two Paths to Agentic Power

Personal Agents

Personal Agents

Personal Agents (Claw level power)

1

Secure Sandbox Execution: Isolated native iOS/Android sandbox for hardware-level privacy and high performance.

2

Native Mobile Superpowers: Direct access to Camera, Photos, GPS, Files, and Contacts with integrated native skills.

3

Evolving Memory & Schedules: Agents learn from your conversations and operate on autopilot with time/location heartbeats.

4

On-Device Edge Models: Support for local LLMs like Gemma 4 and secure storage of MCP/HTTP keys and OAuth.

Remote Agent Hub

Remote Agent Hub

Connect your existing agents

1

Already running Claude Code, Codex, OpenClaw, or other AI tools? Connect them in seconds via MCP, Claude Code Channels, or REST API.

2

Get approval requests, status updates, and agent notifications on your phone. Unblock your coding agents from anywhere — not just your desk.

One app. Every type of agent. Your personal assistants and your remote workforce, side by side. Manage, approve, delegate, and monitor — all from your pocket.


The First Truly Agentic App

Most AI apps are just interfaces for a chat model. HITL is different. Your agents don't just talk — they operate.

  • Environment-Aware: Your agents act within your device's environment, controlling app features and system tools directly.
  • Self-Orchestrating: Your agents manage their own workflows, unblocking themselves and coordinating with other agents.
  • Command & Execute: Tell your Companion what you need. It doesn't just give you a plan; it executes it across your phone and connected services.

Stop chatting with your AI. Start putting it to work.

Agentic App Showcase

Personal Agents: Desktop-Class Power in Your Pocket

Secure Mobile Sandbox
Sandbox

Your Secure Mobile Sandbox

Privacy and security are baked in. Your agents run within the native iOS and Android sandbox, ensuring your data is protected while maintaining high performance.

Mobile Superpowers & Native Skills
Superpowers

Mobile Superpowers & Native Skills

Give your agents direct, permissioned access to your phone's hardware. From Geolocation and Contacts to Files and Camera, your agents interact with the world around you in real-time.

Evolving Memory
Memory

Evolving Memory & Learning

Your agents possess persistent memory, learning from every interaction to become more helpful and personalized over time.

Autonomous Scheduling
Schedules

Autonomous Heartbeats & Scheduling

Set your agents on autopilot. With time and location-based heartbeats, they can trigger workflows, send updates, or perform tasks based on where you are.

MCP Servers Dashboard
Connectivity

Secure Connectivity (MCP & OAuth)

Connect to any external service via MCP or HTTP. HITL handles secure key storage and OAuth management for safe interactions.

Gemma 4 Local AI
Local AI

Local-First Intelligence: Gemma 4

For supported devices, run state-of-the-art Gemma 4 models directly on-device. Lightning-fast responses and maximum privacy for sensitive data.

Remote Agentic Control

OpenClaw Integration
OpenClaw

OpenClaw Native

Access your remote coding agents, OpenClaw, and complex agentic workflows through a single, central communication hub.

Claude Code Agent Management
Claude Code

Claude Code Channels

First-class support for Claude Code sessions. Bi-directional communication lets you approve code changes, review PRs, and guide your Claude Code agent — all from your pocket.

MCP Approval Centre
MCP Hub

Any Agent. One Hub.

Any coding agent or desktop/cloud agent that supports MCP can connect — Gemini CLI, Codex, Cursor, and more. Manage all your HITL requests and iterative planning from one central hub.

Smart Choice Buttons
Smart Choices

One-Tap Decisions

Agents present smart choice buttons so you can approve, reject, or redirect with a single tap. No typing required — just decide and go.

Dashboard Active Messages
Dashboard

Command Centre View

See every agent's pending requests at a glance. Your dashboard shows active approval requests, status updates, and waiting agents — all in real time.

Live Agent Inbox Feed
Inbox

Live Agent Feed

View all your active messages and historical interactions from all your connected agents.

Claw-Level Power.
Mobile-Level Safety.

Near feature parity with desktop agents, but built for the pocket. Trade the terminal for the vault.

Persistent Memory

Persistent Memory

Agents extract facts and learn your preferences. View and edit exactly what they know about you in the secure vault.

Smart Schedules

Smart Schedules

Automate your life with triggers for **Time** and **Location**. Native OS workflows that react as you move through the world.

Dynamic Learning

Dynamic Skills

Your agents don't just use tools—they learn them. As they encounter new tasks, they evolve and acquire capabilities with your approval.

HITL vs. Alternatives

FeatureHITL (Mobile Sandbox)Server Agents (Openclaw/Hermes/Nanoclaw)Cloud LLMs
Security ModelHardware-Level SandboxSoftware ContainerCloud Managed
Native SensesCamera, GPS, Health, FilesNone / MockedText / Image Only
On-Device MemoryPersistent & PrivateLocal DBData Harvested
Smart SchedulesOS WorkflowsServer CronLimited

Private by Default. Secure by Design.

Your agents, your preferences, your keys. They live on your phone — not our servers.

On-Device by Default

Your personal agents, preferences, and conversation history live only on your phone. Nothing leaves your device unless you choose to connect remote agents.

Encrypted Relay

Remote agent communications go through a relay that can optionally be end-to-end encrypted. You control the encryption — we never see your messages.

Bring Your Own Keys

Your API keys are stored locally in secure storage on your device. Full cost control with your own LLM providers — no markup, no middleman.


Meet Your New Companion.

Join the waitlist. Get early access to your personal agent squad and the remote agent hub.

Loading form...

Frequently Asked Questions

What is HITL?

HITL is the first truly agentic mobile operating system app. It allows you to deploy personal AI agents with native mobile superpowers (GPS, Files, Contacts) and serves as a central hub to manage your remote coding agents and workflows.

What platforms are supported?

HITL is available on both iOS and Android. While we strive for complete feature parity, iOS has native restrictions regarding local notification reading.

Are local models supported?

Yes. Support for on-device edge models (like Gemma 4) is rolling out for supported devices, enabling maximum privacy and offline intelligence.

How is HITL different from OpenClaw or Hermes?

Unlike desktop-first tools, HITL is built for your phone. It requires no PC or Mac for setup—it’s a straightforward, mobile-native experience. While HITL isn't designed for long-running, heavy coding execution on-device, it acts as the perfect mobile interface to guide and communicate those agents remotely.

Which LLM providers are supported?

HITL natively supports Gemini models. However, we strongly recommend using Bring Your Own Key (BYOK) which grants access to a wider selection of state-of-the-art models (Anthropic, OpenAI, OpenRouter) and ensures maximum privacy for your interactions.

Is HITL free?

HITL offers a robust free plan. To maintain high performance, there are limits on interactions via our secure relay server, app-provided LLM tokens, and the total number of active agents in your squad.

Is my personal data secure?

Privacy is our priority. All secrets, memories, files, and conversation histories are stored strictly on-device. Local agent definitions remain private unless you explicitly enable cloud sync. While remote interactions via MCP/API use our secure relay, other remote paths like Claude Code channels and OpenClaw are established as direct, secure connections.