ContactBook a demo

Agent identity

Every agent is an identity.Not a credential with a prompt.

An agent that authenticates with a borrowed token has no identity of its own, only someone else's authority. Trustivan models each agent as a first-class identity: who owns it, how autonomously it acts, which credentials it holds, whose authority a provider reports it borrowing, which tools it is entitled to and whether it is still live.

AvailableAgent identity, ownership, autonomy, credentials and lifecycle · no connector discovers an agent, and no screen registers one

Book a demoRead Identity ≠ Credential
support-copilotFirst-class agent identity · not a shared service account
AgentOwnedScoped
Identity
agent:support-copilot · service principal sp-7f3e
Owner
Customer Platform team · assigned by M. Okafor
Autonomy
Human in the loop · operator declared
On behalf of
Not reported by any provider
Credentials
2 bound · runtime credential (expires in 30 days) · AWS access key
Entitlements
6 tools declared by an operator · 1 with no declared effect
Lifecycle
Active · last seen 2 hours ago · certification due in 23 days
Identity ≠ credential: the record survives rotation; the credential does not.Illustrative · synthetic demo estate

The problem

Borrowed authority has no owner.

When an agent acts through someone else's credential, every question that matters in an incident (who is this, who owns it, what may it do) has the wrong answer.

Developer Tokens

An agent running on its creator's personal token carries their entire access and outlives their involvement

Shared Service Accounts

Five agents on one service account means five purposes, five owners and no way to revoke just one of them

Unscoped OAuth Grants

A grant consented to in a hurry gives an agent every scope the app asked for, with no expiry and no review

The model

The relationship model, extended to agents

An agent sits in the same chain as any non-human identity: owned by a human, holding credentials, authenticating to services, reaching resources, performing actions. What it adds is the tools an operator entitled it to call. Agent-to-agent delegation and the data an agent reads are not recorded, by design.

What you get

Six things every agent record carries

The record is the unit of governance. Credentials rotate, prompts change and tools come and go; the identity, its owner and its decisions persist.

See agent provenance

Agent Service Principals

An identity of the agent's own, so logs, grants and findings point at the agent rather than at its creator

Owner And Autonomy

An accountable owner, read from provider tags or assigned by a person, and an autonomy level declared from human-approves-each-action to autonomous

Bound Credentials

Every credential the agent holds mapped to its record, with age, exposure and rotation status

Tool Entitlements

The tools an operator declared this agent may call, enforced at the runtime gate rather than written into a prompt

Lifecycle

Active, stale, orphaned, quarantined and decommissioned as explicit states; a quarantined or decommissioned agent is denied every action

Origin

Where an agent came from is reported as unknown when nothing recorded it, rather than reconstructed from a guess

Beyond the service principal

On behalf of whom, and for how long

Agent identity has two dimensions service accounts never had: the identity an agent acts for, and a lifetime that may be a single task.

Acting On Behalf Of

Where a provider reports whose authority an agent borrows, the record keeps it apart from ownership, and an agent acting for another agent is a finding. No rule caps an action by the requester's authority

Bounded Runtime Credentials

An agent authenticates to the gate with a credential a person mints for one to ninety days, shown once, stored as a digest and revocable. Brokering a credential per task is a future capability

Agent-Created Identities

Agents that provision identities or other agents leave no lineage any connector reports, so origin is shown as unknown. Inventing that edge is not built, by design

Lifecycle

Enter. Own. Review. Retire.

Most agents are experiments that either become permanent or get forgotten. The lifecycle makes both outcomes explicit, with an owner accountable for each transition.

  1. Enter

    An agent enters the inventory only through a connector sync, and no connector reports one today; the demo estate is seeded. No screen registers an agent.

  2. Own

    Ownership is read from provider tags or assigned by a person. An agent with no accountable owner, or no declared autonomy, is a finding.

  3. Review

    Certification campaigns ask owners to confirm access, and entitlement drift reports declared tools going unused. Both have an API and no console screen yet.

  4. Retire

    Quarantine or decommission the agent in the platform and the gate denies every action it attempts. Revoking its credentials at the provider is still done there.

Explore ownership and lifecycle

AI Agent Security

Continue through the pillar

Identity is the anchor. Tools, permissions and runtime decisions all attach to the agent record.

Explore agent use cases

Decide what an agent may do before it acts.

See how TRUSTIVAN connects identity, credential, access, agent and action context into one control plane.

Book a demo