Four concepts cover most of what you need to know.
Agent
A pre-trained AI model specialized for a specific enterprise task. Agents are stateless and process one operation at a time. Each agent belongs to exactly one workspace and has a lifecycle status: deploying, active, paused, or error.
Operation
A single unit of work performed by an agent — one document reviewed, one invoice processed, one ticket triaged. Operations are the unit of billing: plans include a monthly operation quota, and billable operations count against it.
Confidence threshold
A configurable score below which an agent will escalate to human review instead of acting autonomously. Higher thresholds mean more escalations but fewer autonomous mistakes; most teams start around 0.85 and tune down as trust builds.
Integration
A connection between Anysola and an external system (ERP, CRM, ticketing, storage) through which agents read inputs and write outputs. Integrations are managed per-workspace under Dashboard → Integrations.
How they fit together
A deployed agent receives work through an integration, performs operations, and escalates anything below its confidence threshold to a human queue.