Type what you want in plain language: "answer my Telegram overnight and send a morning digest". The builder picks the right engine, deploys it on its own micro-VM, connects your channels, and streams everything it does to a live dashboard.
Three steps. No YAML, no Docker, no server.
A WhatsApp bot for your shop's enquiries. A coding agent you can message from Telegram. Support-email triage for your team. Plain language is the whole interface.
The builder recommends the best of 13 hostable engines for the job, provisions an isolated micro-VM, seals your secrets with encryption, and connects your channels.
A live activity feed, per-agent spend, message counts, and a fleet view. It's ClawMetry: knowing what your agents are doing is the point, not an add-on.
The parts you'd otherwise wire together yourself, included on day one.
An AI engineer reads your description and recommends the right runtime, chat-first, coding-first, or multi-agent. Override it any time.
Telegram, WhatsApp, Slack, Discord, Signal, email, iMessage, web chat and more, depending on the engine you pick.
Every agent gets its own micro-VM on its own private network. Secrets are sealed end to end; the infrastructure API never sees them in plaintext.
Start on the included Claude model with pay-per-use pricing, or bring your own Anthropic or OpenAI key. Switch either way, per agent, any time.
Activity timeline, token spend tied to each task, and cost per agent per day. You watch every move, in real time.
Every agent gets a built-in web chat before you connect a single channel, so you can test it seconds after it boots.
Outgrew an engine? Move the agent and its context to a new one. The builder compiles a handoff briefing so the new agent knows everything the old one did.
All your agents on one screen: status, plan, spend today. Retire, restart or rebuild from one place.
If your use case needs more than the builder covers, book a call and we'll build it with you.
The builder hosts the agent runtime that fits, so you don't have to pick a framework before you've even started.
A dedicated VM per agent, priced like a coffee. Model usage is pay as you go, with no markup when you bring your own key.
1 GB micro-VM. One-shot coding agents, PicoClaw, Hermes, Claude Code, Codex and more.
2 GB micro-VM for the heavier engines: OpenClaw and NanoClaw.
No. Exploring and designing your agent is free, and agents can launch on the included model with pay-per-use pricing. You can switch any agent to your own Anthropic or OpenAI key later.
Each agent runs on its own isolated micro-VM with its own private network. Your tokens and keys are sealed with end-to-end encryption before they ever leave the control plane.
Channels depend on the engine: Telegram, WhatsApp, Slack, Discord, Signal, email, iMessage, SMS, Matrix, IRC, LINE and more. Every agent also gets a built-in web chat.
Observability. ClawMetry started as a real-time dashboard for AI agents, so every hosted agent comes with a live view of what it's doing and what it's spending. You describe, we build and host, you watch every move.
Talk to us. The team behind the builder also does custom agent builds; the founder call link is inside the product.
Describe it in one sentence. Watch it run minutes later.
Open the Agent Builder →