New: Kimi K3 is live — Moonshot's 1M-context flagship, on Claude Code, Hermes & OpenClawLaunch one
Long-horizon agents, everywhere

Any harness, any LLM — cloud-hosted agents on demand.

No Mac mini, no setup. Long-horizon agents, reachable from any channel.

Starts at $3/mo — parked agents are free.

Zero Data Retention (ZDR) supported

Omnichannel

Reach your agents anywhere.

Talk to the same long-running agent wherever work starts. Its history, tools, and state follow it across every connected channel.

WhatsAppiMessageTelegramSlackWebA2A protocolCLI

One-click launch

Launch your agent faster than ordering a burger.

No lock-in on harness or model — you have full control. Upgrade to a cooler setup anytime.

Every harness and model we support

Claude CodeCodexHermesOpenClawGeminiDeepSeekZ.aiKimi K3

Clone to cloud

Clone your local agent to the cloud.

Never start over. Keep everything you've built up locally on your laptop and continue your work in the cloud — with no downtime.

Clone your agent →

Durable runtime

Long-horizon agents, managed for you.

Every agent runs as a managed worker with durable state, full timeline, snapshots, backups, and restore handled by Asteroids.

Infinitely long continuous history

Keep instructions, files, decisions, and conversation history across long-running work.

Managed recovery

Snapshots and backups let the agent restore after faults without losing the thread.

Secure sandbox

Isolated environments boot in seconds with only the tools you attach.

Clone from local

Paste one prompt into your local agent and it clones itself to the cloud — same instructions, model, and MCP servers.

Scale and park

Spin up more agents in seconds. Park idle agents for free until work returns.

Self-evolving (coming soon)

Your agent distills what it learns into an ever-sharper version of itself.

Developer-first

Built for developers. Your whole fleet, one command away.

Asteroids is developer-first: everything the UI does is also a CLI command and a REST API. Create agents, send messages, stream replies, swap models — scriptable from your terminal, callable from your code, ready to wire into CI or your own product.

Read the docs →See what developers build →

FAQ

Questions, answered

The short version of how Asteroids works — harnesses and models, channels, pricing, and what happens to your agent's memory.

What is Asteroids?

Asteroids is a managed agent as a service. One click launches a long-lived, always-on AI agent that runs in a managed sandbox, keeps its history, and can be reached wherever you work — no Mac mini, no setup, no ops.

Which agent harnesses and models can I run?

Any of them. Pick a harness — Claude Code, Codex, Hermes, or OpenClaw — and a model, and launch. You can swap the model or harness later without losing the thread: the agent keeps its full history across the change.

I already run Claude Code locally — can I clone it to the cloud?

Yes. Clone to cloud turns the Claude Code, Codex, or OpenClaw you already run into an always-on cloud agent: paste one prompt into it (or run ast clone from the terminal) and it recreates itself on Asteroids with the same instructions, model, and MCP servers. It copies config only — your secrets, API keys, and session history never leave your machine — and re-running it syncs your latest instructions to the cloud agent.

Do I need to write code or paste API keys?

No. Launching an agent is one click — there is no SDK to install, no API key or bot token to paste, and no infrastructure to stand up first. Platform plumbing stays out of your way.

Which channels can I reach my agent on?

The same agent answers on WhatsApp, iMessage, Telegram, Slack, Discord, web chat, the A2A protocol, and the CLI. Its history, tools, and state follow it across every connected channel.

How much does it cost?

Agents start at $3/mo each plus usage. Parked agents are free, so you can keep an agent around without paying for it until you need it again. Signing up grants a $3 credit with no card required, and the full metering breakdown is at /pricing.

Is my data retained?

Asteroids supports Zero Data Retention (ZDR).

What happens when my agent restarts — does it lose its memory?

No. Agents keep an infinitely long continuous history, and managed recovery — snapshots, backups, and restore — brings an agent back after a fault or restart without losing the thread.

Can I switch models or harnesses later?

Yes, anytime — even mid-task. There is no lock-in on harness or model: change either and the agent and its history stay yours, so you can follow the frontier as it moves.

How is Asteroids different from running a bot myself or using a developer platform?

A DIY bot sleeps, forgets on restart, and makes you the ops team. Developer platforms (like Claude Managed Agents or OpenRouter Spawn) hand you a toolkit and no channels. Cheap single-framework hosts lock you to one framework with no Slack. Asteroids gives you every harness, every channel, and managed durability in one click. See the full comparison at /compare.

No lock-in. Full control.

One click to launch. Zero lock-in after.

Swap harness, model, and channels whenever you like — the agent and its history stay yours. Parked agents are free.

Launch one