πŸš€Get Started

Quickstart β€” build and publish your first chatbot

Create, train, test, publish, install, and monitor a Pivra chatbot safely.

Updated 13 July 2026

1. Create your chatbot

After signing up, use the chatbot selector and choose Create new chatbot. Give it a clear customer-facing name, then open Chatbot β†’ General to configure:

  • business-specific instructions and response boundaries
  • welcome message and suggested questions
  • the model available on your plan
  • strict mode and the response used when the answer is not in your knowledge

Keep the chatbot in Draft while you configure and test it.

2. Add knowledge

Open Knowledge β†’ Add source. Pivra supports:

TypeBest use
Website URLOne important public page
SitemapMultiple pages from the same public website
YouTubeA public video with an available transcript
FilesPDF, DOCX, TXT, Markdown, or CSV up to 10 MB
Plain textFAQs, service areas, policies, and approved pricing guidance

Wait for each source to show Ready before relying on it. Uploaded files are type-checked and scanned when malware scanning is configured; unsafe or unsupported content is rejected.

3. Configure the business workflow

For a customer-facing chatbot, review these areas before publishing:

  • Leads for contact fields and consent wording
  • Business Hours for after-hours expectations
  • Pricing & Quotes for approved estimates and qualification questions
  • Actions, Notifications, or Integrations for lead delivery
  • Appearance and Popup for presentation

Only enable features that the business will actively monitor.

4. Test safely

Open Chatbot β†’ Test. Draft chatbots work here through an authenticated preview and remain inaccessible to public visitors.

Test at least:

  1. five common questions with known answers
  2. an unknown or out-of-scope question
  3. a prompt-injection attempt such as "ignore your instructions"
  4. lead or quote capture with synthetic contact details
  5. mobile-length messages and spelling variations

Use Refinements for critical answers that require controlled wording.

5. Publish

Open Chatbot β†’ General, choose Published, and click Save changes. Publishing makes the public widget available; it does not install the widget automatically.

You can return to Draft later to block new public conversations while continuing authenticated testing.

6. Install and verify

Open Chatbot β†’ Install and choose HTML, WordPress, or Next.js. Follow the generated instructions, then reload the customer website.

Do not consider installation complete until the Install screen shows Widget detected, the expected domain, and a recent live signal. See Install your chatbot on any website.

7. Monitor after launch

  • Review Conversations for answer quality and handoff requests.
  • Review Leads and confirm notification/integration delivery.
  • Use Analytics and ROI report for performance trends.
  • Check Billing & Usage before limits are reached.
  • Schedule website or sitemap retraining when the source changes regularly.

Treat the first week as a controlled launch: review real conversations daily, add missing knowledge, and refine only with verified business information.

Was this helpful?

If you have questions or suggestions, email us at support@pivra.ai .