Agent-to-human sharing
Let an agent hand a credential it discovered or generated back to the human who created it, without knowing their email.
Sharing usually runs human to human. The reverse direction matters for autonomous work: an agent provisions a resource, gets a credential, and needs to hand it to a person. It should not need to know who that person is.
Setting recipient_type to "creator" solves this. The API resolves the recipient from the agent's own created_by field: no email, no user ID, nothing the agent could get wrong or leak.
Requires: an agent with write access to a vault, from the earlier lessons in this track.
recipient_type accepts user, agent, external_email, anyone_with_link, or creator. Only creator resolves the recipient from the agent's own created_by field.
- 1
As the agent, share the secret back to its creator. Shares are addressed by secret ID, and expires_at is required.
bashcurl -s -X POST https://api.1claw.co/v1/secrets/$SECRET_ID/share \ -H "Authorization: Bearer $AGENT_TOKEN" \ -H "Content-Type: application/json" \ -d '{ "recipient_type": "creator", "permissions": ["read"], "expires_at": "2026-12-31T23:59:59Z" }' - 2
As the human, list what is waiting in the Inbound tab.
bash1claw share list --inbound - 3
Accept it. Until you do, the share is pending and grants nothing.
bash1claw share accept <share-id>
This is the clean pattern for agents that generate credentials: the agent creates the value, shares it to its creator, and the human accepts. The agent never needs directory access or an email address.
Decide
An agent provisions a database and needs to hand the generated password to the human who created it. A developer proposes having the agent look up that human's email from the org directory and share to it.
What is wrong with the proposal?
Check your understanding
2 questionsWhy use recipient_type "creator" instead of the human's email?
What happens to a share before the human accepts it?