Skip to main content

​
Purpose

The Agent Library is the platform’s compounding asset. Every agent added is available to every application above it, under one governance model — so the cost of the hundredth agent is a fraction of the cost of the first. This page states direction and sequence, not dates.

​
Where the library stands

Exterview’s 26 agents are the library today, each labelled Available or Upcoming. Dotted lines are Upcoming growth.

​
The seven business functions

An agent’s address on the platform is its business function, its stage within that function, and the industry its expertise is tuned for. The function is the horizontal axis, and it is what makes 100 reachable without a second platform. A function with no agent registered yet is a real, empty slot: the platform recognizes it, filters for it, and shows it as pending registration. No agent is invented to fill it and no code is assigned locally — registration goes through the same catalog every existing agent passed.

​
What growth actually costs

The first application carried the platform: identity, permissions, approvals, evidence, integration, evaluation and the regulatory posture. That work is done and it is shared.

Governance — reused

A new agent inherits the authority model, the approval gates, the record and the fairness measurement. None of it is rebuilt.

Integration — reused

Identity, calendar, meetings, mail and system-of-record connections are platform services.

Expertise — new each time

What is genuinely new per agent is the domain judgment: what counts as evidence, and what good looks like. That is the part worth paying for.

​
What becomes an agent

Not every capability deserves one. The test we apply:

It is accountable for one job

If you cannot write its mandate in one sentence, it is either a feature of an existing agent or two agents wearing one name.

It produces a judgment backed by evidence

If the output is a record rather than a judgment, it belongs in a system of record, not in an agent.

It can be governed

If we cannot state its authority, its approval points and what it may never do, it is not ready to be built.

A customer has asked for it

Customer demand is the primary input to the sequence. Speculative agents are how a library becomes shelfware.

​
How the sequence is decided

1

Demand

What customers are asking for, weighted by how many are asking and how much of their day it consumes.
2

Evidence readiness

Whether the platform already holds the evidence the agent would need. An agent without evidence is a guess with a job title.
3

Governability

Whether the authority, approvals and prohibitions can be stated clearly enough to be enforced.
4

Reuse

Whether it extends the shared library or only serves one workload. Shared capability wins.

​
What we are not building

The library is a curated set of governed, pre-built agents. It is deliberately not an open marketplace, a plugin ecosystem, or a hosting service for third-party agents. An agent that makes judgments about people carries obligations that cannot be delegated to an unknown publisher. Every agent in the library is one we built, tested and stand behind. Organizations needing something bespoke get it as a services engagement, built to the same standard.