Plain code first, AI where it helps
Most of an automation is ordinary code: fetching, filtering, comparing and sending. Your assistant writes those parts without AI, because rules are faster, cheaper and always give the same answer.
For the one step that needs reading comprehension, the script calls ClickRuntime's AI with a short prompt and gets text or structured JSON back. There is no separate AI account to set up and no key to paste anywhere.
- Label support emails, form replies or tickets and route them.
- Summarise a day of channel messages into three lines.
- Pull names, amounts and dates out of free text into a Google Sheet.
- Decide whether a changed web page is a meaningful change before alerting you.
Small prompts, predictable usage
Every plan includes a monthly number of AI tokens, including on the free plan: input tokens are the text a script sends and output tokens are what the model returns. Scripts trim and batch what they send, ask for short answers, and use structured output, so a typical label costs a few hundred tokens.
Before each call ClickRuntime counts the prompt. A call that would go over the month's input tokens is not made, and the script can fall back to a rule or skip the item. You can see the tokens each run used in its run history and the month's total on the Billing page.
Bring your own model if you prefer
If you already use OpenAI, Anthropic or another provider, save its key as a secret in the ClickRuntime dashboard and the script can call that provider directly. The free plan works this way too.