Trained on what you already wrote
Point it at your help centre, sitemap, PDFs or past resolved conversations. It re-indexes when the source changes, so answers don't drift from your docs.
Chattypie's agent reads your help centre, past conversations and product data, then handles the repeat questions end to end. It hands the rest to your team with the context already gathered.
Free plan foreverNo credit card requiredLive in under 5 minutes

An agent that can only paraphrase your FAQ deflects nothing. Chattypie's agent looks things up, takes actions, and knows the point at which it should stop.
Point it at your help centre, sitemap, PDFs or past resolved conversations. It re-indexes when the source changes, so answers don't drift from your docs.
Look up an order, check a subscription, trigger a refund, call your own API. The agent runs the step and reports what happened in the thread.
Scope what it may discuss, what it must never promise, and the confidence floor below which it stops answering and fetches a person instead.
When it escalates, your teammate opens a thread that already has the summary, the account, and what the agent tried. No re-reading from the top.
On the tickets it doesn't own, the agent still drafts, rewrites, shortens and translates. Your team edits and sends.
Resolution and escalation rates over any date range, so you can see how much the agent actually closes rather than how many conversations it touched.
There is no prompt engineering step and no rebuild of your content. Give it the URLs and files you already maintain, and it indexes them.

The useful number is not how many conversations the agent touched. It is how many it closed on its own, and how often it handed over instead.

The whole thing is a single afternoon, and the free plan doesn't expire while you evaluate it.
Paste your help centre URL or sitemap and let it crawl. Add files and past conversations if you have them.
Choose the tone, the topics it owns, the things it must never say, and when it should fetch a human.
One agent across web chat, email and WhatsApp. Same knowledge, same guardrails, one inbox behind it.
It says so and escalates rather than guessing. You set the confidence floor, and every escalation arrives with a summary of what the customer asked and what the agent already tried.
Answers are grounded in the sources you give it rather than the open web. You can scope it to refuse anything outside those sources, and review the ones it got wrong in the conversation log.
Yes. It can look up an order, check a subscription, or call your own API and report the result in the thread. You choose which actions it may take and which need a human to approve.
Each plan includes a number of agents, from 2 on the free plan to 50 on Business, and extra agents are $3 per agent per month. There is a free plan with no card required and no expiry.
No. Chattypie is hosted by us, which keeps the agent, the crawler and the model routing on infrastructure we can keep patched.
Free plan, no credit card, and a live agent before your coffee goes cold.