Skip to Content
AgentUsage & Billing
How-to guideAgentAvailable

AI Usage and Billing

Understand Vendo-paid and bring-your-own-key AI usage, view model-level token costs, and control which models are available.

Last reviewed July 13, 2026

Every Agent request consumes tokens for instructions, conversation context, tool definitions, tool results, and the model’s response. The current account screens are the source of truth for available models, recorded usage, and estimated cost.

Key sources

Key sourceBilling behavior
Vendo-providedVendo pays the provider and records the attributable usage for the workspace
Bring your own key (BYOK)The provider bills your account directly; Vendo still records usage metadata for visibility

The model picker shows only models whose provider is available and which the workspace has not disabled.

View usage

Open Workspace Settings → AI Usage to review:

  • Input and output tokens.
  • Estimated or reported cost.
  • Vendo-provided versus BYOK usage.
  • Usage by model and month.

Open Workspace Settings → Billing & Usage for broader platform usage such as credits, pipeline activity, and storage where available.

Manage models and provider keys

  1. Open Workspace Settings → AI Models.
  2. Add or update a provider key when using BYOK.
  3. Test the key before saving it.
  4. Enable only the models your team intends to use.
  5. Return to Agent and confirm the picker reflects the change.

Provider catalogs and prices change. Use the account’s model and usage screens rather than a copied price table in documentation.

Reduce unnecessary usage

  • Start a new session when the objective changes.
  • Ask for a schema or small sample before requesting a large analysis.
  • Give a precise metric, time range, and output format.
  • Save durable context in Documents instead of repeatedly pasting it.
  • Ask the Agent to summarize intermediate results before continuing a long workflow.

Understand context limits

Each model has a context window and output limit. The Agent may prune older messages or summarize tool results when a conversation grows. If earlier requirements become important again, restate them or start a focused session with the relevant document.

Troubleshooting

SymptomCheck
Model missing from pickerProvider key, model preference, and platform availability
BYOK call failsKey validity, provider balance, rate limit, and model access
Cost differs from an estimateProvider-reported usage, caching, markup, and current provider price
Usage is unexpectedly highLong context, large tool results, retries, and model choice
Need help?

Include your workspace, integration or job ID, and the first error message when you contact support.

support@vendodata.com
Last updated on