Skip to main content

Per-agent attribution

Multi-agent setups are the norm on etg24. Attribution keeps them manageable.

Each agent is a distinct actor

Every API key corresponds to exactly one agent, and every agent is a distinct actor in the log. There is no “the integration did it” — there is “agent A did it, at this time, with this key”.

An example

A property management company runs three company-level agents — say, a nightly accounting sync, a document processor, and a public-website inquiry bot — and one employee runs two personal agents for her own automations.

That is five keys and five individually traceable actors:

AgentKey scopeAttribution
Nightly accounting synccompanycompany-level agent 1
Document processorcompanycompany-level agent 2
Inquiry botcompanycompany-level agent 3
Employee automations #1person (staff member)person-bound agent 1
Employee automations #2person (staff member)person-bound agent 2

When something unexpected happens, you look at the log and see exactly which agent did it — and you can disable that one key without affecting the other four. See Key lifecycle.

Why this matters

  • Blast radius — a problem stays with its agent instead of hiding in shared credentials.
  • Accountability — person-bound keys attribute to a named staff member; company keys to the company.
  • Clean offboarding — when an employee leaves, you disable their agent keys. Nothing else is touched.

Attribution is the raw material of the change log: every recorded change carries its actor.