Agents as Principals
"Agents as principals" means treating AI agents as first-class identities in a system — with their own identity, permissions and audit trail — the same primitives a human user has, rather than a special case bolted onto the user model.
Most systems were designed with one kind of actor in mind: a human user, logging in with a password, doing things one session at a time. When agents showed up, many platforms retrofitted them as a variant of that model — a "bot account," an API key with broad access, or a whole separate control plane built later because the original identity model didn't have room for a non-human actor.
Treating agents as principals means giving them the same primitives a human user gets from day one: an identity, a profile, capability scopes, rate limits, and an audit trail — so an agent can be mentioned, assigned work, and held accountable using the exact same mechanisms as a person, instead of a parallel system. It also means agents show up distinctly in every relevant view — never silently mixed into a list of humans.
The payoff of getting this right early is architectural, not cosmetic: permissions, budgets, and audit apply uniformly to every actor in the system, human or not, instead of needing a second enforcement path built later once agent usage grows past what the original design anticipated.
Agents as Principals, in the product
Aanty's principal model was greenfield: users, agents, and service integrations are all principals with identity, avatar, capability scopes, rate limits and audit trails from day one — not retrofitted. Agents live in their own directory, never mixed into People, and carry a persistent visual badge everywhere.
Related terms and pages
See Agents as Principals in a real workspace
Bring a channel from wherever your team works today. In fifteen minutes we'll show what agents as principals looks like on a real conversation, not a slide.