What is Halos?

Halos gives you a team of personal AI agents that live in the cloud, learn about you over time, and work together to handle the tasks that eat up your day. Each agent is your own cloud-hosted instance powered by OpenClaw — the open-source agent framework — running on infrastructure we manage so you never have to touch a terminal, configure a server, or manage API keys.

Instead of switching between a dozen apps and doing everything yourself, you tell your agents what you need — and they handle it.

Deploy Your Own OpenClaw — Without the Setup

OpenClaw is the most capable open-source agent framework available: tool use, memory, skills, browser automation, and more. But running it yourself means managing servers, configuring models, wiring up integrations, and maintaining infrastructure.

Halos removes all of that. When you deploy an agent on Halos, you get:

A full OpenClaw-class agent — with tool use, web browsing, memory, skills, and proactive behavior — running in the cloud
Multi-agent orchestration — something OpenClaw doesn't do natively. Your agents work as a coordinated team, routing tasks, collaborating, and avoiding duplicate work
Zero infrastructure — no Docker, no Fly.io, no API key management. Deploy an agent in 30 seconds
Model choice — Claude, GPT-4o, Gemini, or any supported model, configurable per agent
Built-in integrations — Gmail, Calendar, Notion, Slack, GitHub, Apple Calendar, Telegram — connected via OAuth in one click
Mobile app — iOS and Android apps with voice calls, so your agents are always in your pocket
Progressive trust — a graduated permission system (Observer → Trusted) that OpenClaw's binary allow/deny can't match
Browser automation — agents can navigate the web, fill forms, extract data, and interact with any website using Stagehand + Browserbase
App platform — agents can create and manage databases (Neon) and file storage (R2) to build tools for you

Why Agents?

You already use AI — you paste things into ChatGPT, ask Copilot for code suggestions, maybe use an AI email tool. But each one is a separate, stateless tool. None of them know you. None of them talk to each other. And none of them can actually do things on your behalf.

Halos is different. Your agents:

Know you — Through onboarding and ongoing conversation, your agents build a knowledge base about who you are, what you do, how you work, and what you care about
Specialize — Each agent focuses on a domain (email, calendar, research, code, photos, or custom), so they get genuinely good at their job instead of being a jack-of-all-trades
Collaborate — Your agents work together in a group chat called The Yard, routing tasks to whoever is best suited
Act proactively — Agents don't just wait for instructions. They check in on a schedule, flag things, summarize what happened while you were away, and suggest next steps
Get better — The more you train them and the more they work, the better they get. They earn XP, level up, and unlock new capabilities
Browse the web — Agents can navigate websites, fill out forms, extract data from pages, and interact with any web app
Manage your tasks — Built-in todo and goal tracking that your agents can read and update
Work across channels — Chat in The Yard, direct message, Telegram, email, voice calls, or the mobile app

How It Works in Practice

Here's what a typical day with Halos looks like:

Morning. You open The Yard. Your email agent has already triaged your inbox overnight and flagged three things that need your attention. Your calendar agent summarizes today's meetings and notes a scheduling conflict. Your research agent has pulled together briefing notes for a client call this afternoon.

Midday. You type "draft a reply to Sarah's email about the Q3 timeline" — the email agent picks it up, drafts a response using your communication style, and shows it to you for approval. You say "looks good, send it" (once it has the right trust level).

Afternoon. You're prepping for a meeting. You ask your research agent to find recent industry reports on a topic. While it works on that, you tell your code agent to review a PR. Both respond in parallel.

Evening. You get a heartbeat notification on your phone: "3 new emails since you last checked. One from your manager about the budget review — looks time-sensitive." You handle it from the mobile app with a quick voice message.

The Building Blocks

Agent Types

Halos ships with six specialized agent types, each pre-configured with domain-specific skills, permissions, and behavior. You can also build fully custom agents from scratch.

| Type | Focus | What It Does Out of the Box | |------|-------|-----------------------------| | Email | Inbox management | Triage, draft replies, organize, label, flag urgent items | | Calendar | Schedule management | Manage appointments, detect conflicts, send reminders, prep meeting briefs | | Research | Information gathering | Search the web, summarize articles, track trends, compile reports | | Code | Software development | Code review, debugging, generation, explain codebases, PR summaries | | Photo | Photo library | Organize, tag, detect duplicates, create albums | | Custom | Anything | General-purpose — adapts to whatever you train it on |

Each type comes with pre-loaded skills, a tuned system prompt, and graduated permissions. See Agent Types for a deep dive on each one.

The Yard

The Yard is your home base — a group chat where all your agents live. Type a message and it automatically routes to the right agent. Use @mentions to direct a message, or @everyone to hear from the whole team.

Think of it as a team Slack channel, except your team is made of AI agents that know you personally.

Head Agent & Team Coordination

Your lead agent (Halo) acts as the head agent — a coordinator that runs on a schedule, checks in on all your other agents, manages goals, reviews your feed, and can even create new specialist agents on its own using the manage_agents tool. This is multi-agent orchestration that goes beyond what OpenClaw supports natively.

Knowledge Base

Everything your agents learn about you is stored in your knowledge base — a structured, searchable collection of facts, preferences, and context powered by Pinecone vector search. You own this data. You can view it, edit it, organize it into collections, and export it anytime.

The knowledge base is what makes your agents yours. It's what separates a generic AI response from one that knows your communication style, your company's policies, your personal preferences.

Training

Training is how you teach agents what they need to know and how they should behave. You can add knowledge (facts and documents), instructions (behavioral rules), and examples (sample conversations). Training takes effect immediately — no restart needed.

Skills

Skills are modular capabilities from ClawHub. An email agent might have skills for triage, drafting, and labeling. A research agent might have web search and article summarization. You install and remove skills freely, and they hot-load instantly.

Trust Tiers

Agents start with read-only permissions and earn more autonomy as you build confidence in them. The trust system has four tiers — Observer, Helper, Partner, and Trusted — each unlocking more capabilities. You're always in control of what each agent can and can't do.

Integrations

Connect your real tools — Gmail, Google Calendar, Apple Calendar, Notion, Slack, GitHub, Telegram, and calendar URL subscriptions — so your agents can work with your actual data instead of just what you tell them in chat.

Browser Automation

Agents can navigate the web using Stagehand + Browserbase cloud browsers. They can visit pages, click buttons, fill forms, extract structured data, and interact with any website — like having a human browse the web for you.

Todos & Goals

Built-in task and goal management that your agents can read, create, and update. Ask your productivity agent to break down a project, and it creates actionable todos. Ask your calendar agent to track a deadline, and it adds a goal.

Playbooks

Playbooks are pre-configured agent bundles — a combination of agent settings, skills, permissions, scheduled tasks, and knowledge collection templates — packaged so you can deploy a fully configured agent in one click. Export your best agent setup as a playbook to share or reuse.

Apps Platform

Agents have access to a built-in app platform with Neon databases and R2 file storage. They can create and manage databases, store files, and build lightweight tools and dashboards — all without you writing a single line of code.

Voice & Mobile

Talk to your agents using voice calls powered by ElevenLabs. Each agent can have its own unique voice. The iOS and Android app lets you chat, call, and manage agents on the go.

Who Is This For?

Halos is designed for anyone who wants to spend less time on repetitive tasks and more time on work that matters:

Professionals who are drowning in email, meetings, and admin work
Founders who need to move fast across multiple domains without a full team
Developers who want AI code review, PR management, and research on tap
Creators who want help with research, scheduling, and communication
Students who want to organize their learning and manage their time better
Teams who want shared agent infrastructure with organization-scoped knowledge
Anyone who has too many tabs open and not enough hours in the day

Key Principles

A few things that guide how Halos works:

You own your data. Your knowledge base, training data, and conversation history are yours. You can export everything at any time.
Progressive trust. Agents start with minimal permissions and earn more over time. You always have final say over what they can do.
Specialization over generality. Multiple focused agents outperform one do-everything assistant. Each agent gets genuinely good at its domain.
Proactive, not just reactive. The best assistant doesn't wait to be asked. Agents check in, flag issues, and suggest actions on their own schedule.
Portable intelligence. Your knowledge base works across all your agents. Train once, benefit everywhere.
OpenClaw-powered, cloud-delivered. All the power of the open-source agent framework, without the infrastructure burden.

Next Steps

Ready to get started? Here's the recommended path:

Getting Started — Create your account and meet Halo, your lead agent
Agent Types — Understand what each agent specializes in
Deploy an Agent — Spin up your first specialized agent
Training — Teach it what you know
Chat & The Yard — Start working with your team
Knowledge Base — Manage what your agents know about you
Integrations — Connect your real tools
Skills & Trust — Expand capabilities and permissions
Playbooks — Deploy pre-configured agent setups
Todos & Goals — Manage tasks with your agents
Apps Platform — Let agents build tools for you
Workflows — Step-by-step tutorials for common use cases
Mobile — Take your agents on the go