Shared machines, simple rules

Acceptable use.

A cloud session runs real programs on hardware we pay for, next to other people's sessions. This policy exists so that stays workable. It is part of the terms of service.

Effective September 18, 2026 · Tomas Godoy Pastore · support@murmello.ai

What Murmello's cloud is for

Running your own software development work: building, testing, debugging and editing your own projects, with a coding assistant, continued from your Mac. That is the whole intended use. If what you are doing looks like that, you are fine.

Do not

• Attack or interfere with anything. No denial-of-service traffic, port scanning, intrusion attempts, credential stuffing, spam or bulk unsolicited messaging, from our infrastructure or against it.
• Try to escape the boundary. Do not attempt to break out of your session's container, reach another user's session or data, read the instance metadata service, or escalate privileges on our hosts. Security research against your own session is welcome — tell us what you find at support@murmello.ai rather than exploiting it.
• Use it as general-purpose hosting. Cloud sessions are development sessions, not a place to run a public website, a game server, a bot, a proxy or VPN exit, a file-sharing endpoint, or any long-running production service. Port forwarding exists so you can reach a dev server you are working on, from your own Mac.
• Mine cryptocurrency or run distributed compute. This includes mining, hash cracking, and renting or reselling your session's compute to anyone else.
• Break the law or other people's rights. No infringing content, no malware distribution, no child sexual abuse material, no material that harasses or threatens, nothing that violates export control or sanctions law.
• Store data you are not allowed to put on a third-party service. That includes data you hold under a confidentiality obligation that forbids it. Murmello is not built or certified for regulated health, payment-card or government-classified data, and you should not put such data into it.
• Evade limits. No creating multiple accounts to collect extra trials or allowances, sharing one account between people, or working around concurrency, storage or rate limits.
• Resell the service. Do not offer Murmello's hosted capacity to third parties as your own product.

Automation and agents

Murmello exists to run coding agents, so automated activity is expected and welcome. What is not welcome is automation aimed at consuming as much free or included capacity as possible, or at keeping sessions alive without doing real work. Allowances are metered in wall-clock time on the machine size you chose, and that includes time an agent spends waiting for a model.

You are responsible for what an agent does under your account. An agent acting on your behalf is you, for the purposes of this policy.

Fair capacity

Capacity is shared and finite. We may limit how many sessions run at once, cap the machine size available, queue or refuse a new session when the cluster is full, stop a session that exceeds its allowance or fails to renew its lease, and reclaim idle resources. None of this deletes your files: a session's work is checkpointed before its disk is released.

What we do about problems

Where the situation reasonably allows, we will contact you first and give you a chance to fix it. Where it does not — an active attack, a serious legal risk, or something destabilising the service for other people — we may suspend a session or an account immediately and explain afterwards. Serious or repeated breaches can end your account under the terms of service, and your data export window applies as described there.

If you think we have got it wrong, reply to us at support@murmello.ai. A human — the same one who wrote this — will read it.

Reporting abuse

To report abuse of Murmello's infrastructure, or content hosted on it, write to support@murmello.ai with enough detail to identify what you are reporting. We investigate every report we receive.