Invocation Logs, Tokens & Cost

Every interaction with an agent is recorded. An invocation is a single model call, captured with its token usage and cost — giving you full visibility into how an agent behaves and what it costs.

Per-agent analytics

An agent's published page shows an Analytics panel summarising its activity:

• Total Chats — conversations recorded for the agent.
• Input / Output Tokens — tokens consumed by prompts and generated in responses.
• Cost Estimate — the modelled cost of that usage.

What's logged

For each conversation, AETHER stores the full exchange — messages, tool calls, citations and token counts — so you get end-to-end traceability for debugging, auditing and compliance. The Conversation Explorer lets you open a conversation and inspect what the agent did turn by turn, including which knowledge it cited and which tools it called.

Cost basis

Cost is derived from each call's tokens and the configured per-model rates (see admin → Models). Different modalities — text, cached input, audio and embeddings — are all metered. Roll-ups across agents, teams and the whole organisation live in the Cost and Usage reports.

Tip: Watch output tokens and average cost per conversation when tuning an agent — overly long responses are the most common, easily-fixed cost driver.

Next steps

• Publishing & Environments — promote a tested agent.
• Cost Report — organisation-wide spend analysis.