Beginner · Systems

Agent Capabilities

Visual diagram · (in preparation) · Math · (in preparation) · Worked example · 3 difficulty levels.

TL;DR. The concrete set of actions an agent can perform — the tools, APIs, and side effects available to it.

Technical Definition

The concrete set of actions an agent can perform — the tools, APIs, and side effects available to it.

How it works

An agent's capabilities are defined by the tools it is given: read a file, query a database, send an email, place an order, spawn a sub-agent. Capabilities should be declared explicitly with typed inputs, described side effects, and reversibility markers, because the model's behaviour is shaped as much by which tools are visible as by its instructions. Exposing fewer, better-scoped capabilities per phase reduces both error rate and blast radius.

Related Concepts

• Agent Authorization — Deciding and enforcing what an authenticated agent is allowed to do, at what scale, and under which conditions.
• Capability-Based Agent Authorization — Authorizing agents by handing them unforgeable, narrowly scoped capability tokens rather than checking roles at call time.
• Agent Capability Vocabulary — A shared, standardized naming scheme for the actions agents can offer and request.
• Context-Aware Tool Selection — Exposing and choosing tools dynamically based on the current task phase rather than presenting every tool at every step.