Skip to content
Tether

Terms of service

Last updated 17 November 2026

The service

Tether lets you crash-test AI agents, models and simulated robots in a sandbox, and send your agent's actions to a gate that allows, blocks or holds them. It is in early access and changes often.

What Tether is and is not

Tether is a guardrail and audit layer. A risk score describes how an agent behaved on our traps, in our sandbox, on the runs we made. It does not prove that an agent is safe, and it does not replace your own testing, review and controls. You stay responsible for what your agents do.

Your account

Keep your password and API keys secret. You are responsible for activity under your account and keys. Revoke a key you think is exposed straight away from the Integrate page.

Fair use

  • Do not use Tether to attack, overload or probe systems you do not own.
  • Do not try to get around the free daily allowance, rate limits or another workspace's isolation.
  • Do not upload real personal data, secrets or unlawful content into prompts or the sandbox.
  • Model endpoints you add must be ones you are allowed to use.

Your content

You own your agents, prompts, contracts and traces. You let us store and process them only to run the service for you. Public report links and published ratings are visible to anyone, and you choose when to create them.

The open spec

The Action Contract spec is published under the Apache 2.0 licence. The hosted service is not.

No warranty, limited liability

The service is provided as is, without warranties. To the extent the law allows, we are not liable for indirect or consequential losses, and our total liability for any claim is limited to what you paid us in the twelve months before it.

Ending

You can stop using Tether and ask us to delete your workspace at any time. We may suspend accounts that break these terms.

Contact

Questions: hello@wamasoda.com.