Token scoping is the practice of limiting an authentication token to only the actions, data, and systems it is meant to access. A scoped token might be valid for one API, one user role, one repository, or read-only access, instead of granting broad, reusable privileges. This is a core identity control in modern cloud and application security.
It matters because stolen or leaked tokens are often easier to abuse than passwords: if the token is overprivileged, an attacker can move laterally, call sensitive APIs, or modify infrastructure. In agentic AI systems, scoping is especially important because models may use tokens to browse, query databases, or run tools. Defenders reduce risk by issuing short-lived tokens, separating read and write permissions, applying allowlists, and rotating credentials regularly. Good token scoping turns a compromised credential into a limited problem instead of a full environment breach.