GUIDE · AI AGENTS
NIST's 2026 work describes AI agent systems as capable of planning and taking autonomous actions that can affect real systems. Voyager uses the customer-facing term Worker for a different unit of design: a governed responsibility with explicit context, tools, authority, approvals, escalation, budget and evaluation.
01AI AGENT SYSTEMS
NIST and CAISI's 2026 agent-security work focuses on systems that can plan and take autonomous actions affecting real environments. The practical business question is not only whether the model reasons well, but what external systems and actions it can reach.
02VOYAGER WORKERS
A Voyager Worker is configured around an outcome and responsibility, the context it may use, instructions, tools, triggers, direct-action authority, approval gates, escalation, model and budget policy, outputs and evaluation.
03CONTEXT ACCESS
A Worker may need customer, project or finance context to prepare useful work without having authority to edit every record or trigger every external action.
04EXECUTION AUTHORITY
One Worker can be allowed to prepare a draft, required to request approval before sending it, and prohibited from changing a commercial term. The boundary belongs to the responsibility, not to a global autonomy switch.
05AUDIT AND EVALUATION
NIST guidance emphasizes documentation, oversight and monitoring. Voyager Worker runs record actions and relevant evidence so organizations can review execution, cost, exceptions and outcomes.
06ENGINES
An Engine can combine multiple Workers, deterministic workflows, intelligence, native capabilities, integrations, policies, approval rules and measurement around a substantial outcome.
DIRECT ANSWERS
Get started
Map what the Worker owns, what context it needs, which tools it may use, which actions it can take and where approval or escalation begins.
