AI usage and keys
AI in Pinlyx runs on your own AI provider key. You add the key once, pick a model, and every AI feature uses it: agents, suggested replies, email summaries and the rest. Your provider bills you directly, and the AI usage and cost page estimates what you are spending.
How it works
Section titled “How it works”- No key, no AI. Pinlyx does not supply a model key for replies. Until you save one, AI features do not run and agent runs show as Failed.
- One set of settings for everything. The provider, model and key under Settings › AI assistant are used by every AI feature. A single agent can use a different model under Identity & tone › Model and creativity.
- Each provider key is stored separately, so you can switch providers without pasting the key again.
- Keys are stored encrypted. The field never shows a saved key.
Add your AI key
Section titled “Add your AI key”-
Go to Settings › AI assistant.
-
In the Model row, click Add your key. The AI assistant settings editor opens. Use Back and Next to move between its steps.
-
On Model, pick a Provider and a model. Each model shows its context size and its price per 1M tokens, so you can compare cost before you choose.
-
On API key, paste the key from your provider. The link next to Test opens the provider’s page where you create a key.
-
Click Test. A working key shows Connected with the response time.
-
On Tone, choose a tone (Professional, Friendly, Casual, Playful or Concise), set the Temperature, and add an optional System prompt with instructions every reply should follow.
-
Leave Advanced on auto unless you know you need it.
-
Click Save AI settings. The Model row now shows Key saved.
The editor also has a Cost calculator: enter tokens per request and requests per month to estimate a monthly bill for the selected model.
Read the usage page
Section titled “Read the usage page”Go to Settings › AI usage and cost. Pick a range of 7 days, 30 days or 90 days.
| Part | What it shows |
|---|---|
| Today | Estimated cost today, with calls and tokens in and out |
| Cost | Estimated cost for the range, and on how many days AI ran |
| AI calls | Number of AI calls in the range, with tokens in and out |
| Daily cost | Estimated cost per day, with today highlighted |
| By feature | Which AI features cost the most, with calls and tokens for each |
Days follow your account time zone.
What counts
Section titled “What counts”The page adds up AI calls that record their token use:
- Agent replies, listed by channel, for example Telegram AI, Social inbox AI and Live chat AI
- Alerts, the checks that decide whether an alert fires
- Other AI features that record tokens, such as the Social composer, Outreach and Email AI
Messages on an agent’s Test tab and reply suggestions from AI suggest reply are not on this page, though your provider still charges for them.
Limits per plan
Section titled “Limits per plan”| What | Free | Pro | Business |
|---|---|---|---|
| AI agents, knowledge base, handoff, alerts, reply assist | No | Yes | Yes |
| Number of agents | None | Limited | Unlimited |
| Continuous learning from replies you type | No | No | Yes |
The exact numbers for each plan are on Plans and limits, and prices on pinlyx.com/pricing. Your provider’s own rate limits and quotas apply on top.
Good to know
Section titled “Good to know”- Cost grows with context. Each agent reply carries the persona, every approved knowledge entry and the recent conversation. Shorter knowledge entries mean cheaper replies.
- Deep learning runs cost more. Learn from your chats with Deep or Everything reads more conversations and uses more provider credit.
- Translation has its own key. The DeepL key lives in the same settings, under Auto-translation (DeepL). See Reply assist.
- Every AI action is on record. Settings › AI transparency keeps an audit log of what AI did in your account. Click a row to see the full input and output.
- Removing a key stops AI. Click Clear next to Test, then Save AI settings, to delete a saved key.