AI Models & Credits
No AI key needed
Every agent and chatbot on AgentLane runs on AgentLane’s own managed AI — you don’t need to bring your own OpenAI or Anthropic key, and nothing in the product reads a stored Anthropic key for AI inference anymore. Usage is metered against your plan’s included AI credits instead.
If your agency previously stored an Anthropic key for a BYOK plan, it’s no longer used to run AI — see the note below.
What runs by default
The default model behind your chatbot and catalog agents is a fast, low-cost Google model (Gemini Flash-Lite class). It’s chosen for speed and low cost per response, and it’s fast enough that most conversations don’t feel any different from a larger model.
A small number of stronger models are also available for the chatbot, selectable in Dashboard → Settings → Chatbot:
- A stronger Google model, for more complex conversations, at a moderate credit cost.
- Anthropic Claude models (Haiku, Sonnet, Opus), where enabled — check the model list shown in your own chatbot settings, since which ones are offered can change without a code deploy.
The models actually offered to you are always whatever the chatbot settings page lists — that list is the source of truth, not this page.
How credits are debited
Every AI response debits your plan’s included AI credits (or, once those run out, your purchased credit balance). The number of credits a response costs depends on two things:
- How much text went in and came out — longer prompts and longer answers cost more.
- Which model answered — each model has a credit multiplier. The default, fastest model has the lowest multiplier; a stronger model costs proportionally more credits per response, because it costs AgentLane proportionally more to run.
In practice: switching your chatbot to a stronger model doesn’t change your plan’s included credit allowance, it changes how fast those credits are used. A conversation on the default model might cost a fraction of a credit; the same conversation on the strongest available model costs several times more.
Choosing a stronger model
In Dashboard → Settings → Chatbot, the Model dropdown lists every model currently available to your agency, with the default clearly labeled. Pick a different one and save — the change takes effect on the next conversation; nothing is retried or backfilled for conversations already in progress.
If your agency previously chose a specific model, that choice is respected. Some accounts have had their chatbot model reset to the default fastest model as part of a cost-optimization change — if your chatbot doesn’t seem to be running the model you expect, check this page and re-select it.
Where usage shows up
Dashboard → Billing shows your AI credit usage. The chatbot’s own “Check usage” tool (if enabled for your clients) reports usage grouped by type, so a client can ask their portal chatbot how much they’ve used without leaving the conversation.