Hosted log: taking the first tenants

Everything else on this site runs on your own machines. The one thing you cannot run yourself is a log that is not yours: a transparency log operated by someone who is not the agent's operator, whose key signs the tree heads, so an auditor, a customer, or a regulator can accept a receipt without trusting you.

That log is running at log.agent-custody.dev (status: a probe from machines that are not ours verifies it every ten minutes). It holds no receipts, only their hashes; its keys are published at /.well-known/agent-custody-log.json for verifiers to pin; and it publishes signed checkpoints to a second host, checkpoints.agent-custody.dev, so a rewrite is detectable by someone who was not watching. We run our own receipts on it. Anyone can register a tenant at app.agent-custody.dev and get a log and a first key at once; the first tenants shape retention, jurisdiction, and what the reports should say.

What a tenant gets: a log of their own at /t/<tenant>/, a bearer token per fleet, a log id their tree heads carry, a welcome sheet with the one config line and the two verifier commands, and a dashboard showing appends against the plan, keys, the latest checkpoint, and every action taken on the tenant. What a tenant does not get yet, said plainly: a second independent signer (the witness, issue #6) or a contract with service levels. Read the proof table before repeating any claim about it.

Prefer to run it yourself? The same server is a container, a compose file, and Kubernetes manifests, and everything the hosted one does, yours does too.

Free is ten thousand appends a month; Team is $50 a month for a million; Enterprise is a conversation. The pricing page has the table and what is not promised on any plan.

Everything else, the gateway, the SDK, the ledger, the verifier, stays open source under Apache-2.0 and needs no account. Get started with that today.

For your security team: the security questionnaire, the threat model, and the compliance mapping.