Skip to main content
Short answers to the questions that come up most often, grouped by theme. Each one links to the page that covers it in full. If something here points at a problem you’re actively hitting, jump to Troubleshooting for symptom-by-symptom fixes.

​
Getting started

An agent is an AI chatbot that answers questions using your content. You give it a knowledge base, shape its behavior with instructions, test it, and embed it on your site. See the Quickstart for the end-to-end path.
Open the dashboard at app.a2v2.ai and follow Create an agent. The whole build loop — create, train, shape, test, ship — is laid out in the Quickstart.
You sign in with email (a one-time passcode is sent to your inbox) or with Google. Your email is your account identity. See Profile.
Your account is your personal login. An organization is the shared workspace that holds your agents, knowledge, contacts, billing, and team. You can be a member of an organization with a role that controls what you can do. More in Core concepts.
No. Building and training an agent is entirely no-code. You only touch HTML when you embed the agent on your site — and even then the simplest option, a Share Link, needs no code at all.

​
Agents & knowledge

Five source types: Text, Files (PDF, DOCX, TXT), Websites, YouTube transcripts, and Q&A pairs. Mix as many as you need. See the Knowledge Base overview.
Adding a source kicks off background processing — the content is read, chunked, and indexed. Only Completed sources are used in answers. A source that’s still Processing or that Failed is invisible to the agent until it finishes. See Manage & reprocess sources.
Reprocess the source from its Actions menu (⋯). Reprocessing re-reads, re-chunks, and re-indexes the latest content. For a website or YouTube source it re-fetches the current content at the URL.
Instructions are the rules that control how your agent behaves — its persona, tone, scope, and how it handles questions it can’t answer. Knowledge tells the agent what it knows; instructions tell it how to act. Every save is stored as a version you can roll back to. See Instructions.
Pick a model in the Sandbox — each option shows its credit cost per message and any HIPAA/Premium badges. Lighter models cost less per message and answer well from a strong knowledge base; premium models cost more. Compare options in the Model reference.
Temperature controls how focused versus creative answers are, on a slider from Reserved to Creative (default 0.5). Lower keeps answers tight and predictable; higher allows more variation. See Tune temperature.

​
Answers & quality

On every question, the agent searches your indexed sources for the most relevant passages and writes an answer grounded in them — this is retrieval-augmented generation (RAG). See How answers are generated.
Almost always the source isn’t Completed, or the wording doesn’t match. Confirm the source finished in the Knowledge Base, then add a closer-matching Q&A pair. Full walkthrough in Troubleshooting.
Add a Q&A pair — a question paired with the exact answer you want returned. Q&A is the most precise way to control a response word-for-word, ideal for pricing, policies, and FAQs.
Sharpen your instructions (persona, tone, formatting) and consider lowering temperature. Test each change in the Sandbox.
Chat with your agent in the Sandbox, and compare two full configurations side-by-side in the Playground. Testing there is for your own iteration. Playground testing is free.
It’s designed to answer from your sources, and you can instruct it to say so when it can’t find an answer rather than guess — telling it not to guess is one of the highest-impact rules you can add. See Fallback behavior.

​
Embedding

Three options on the Embed/Install screen: a Share Link (no code), a floating chat bubble, or an inline iframe. Each snippet is pre-filled for your agent — copy it directly from the screen.
Use a Share Link to send the agent as a URL, the chat bubble for a site-wide assistant on every page, and the inline iframe to place the chat inside a specific page or layout. See Embed & install.
Check the snippet placement (chatbotId present, script not stripped) and clear any page or CDN cache. Step-by-step in Troubleshooting.
No. The widget loads inside an isolated container (a Shadow DOM), so it won’t affect your site’s styles and won’t be affected by them. See Embed & install.
Yes. On Visibility & access you choose Private (only you) or Public, and for public agents you pick Use for Free, Restrict (only your CRM users), or Pay for Access.

​
Credits & billing

Credits are how A2V2.ai meters usage. Instead of counting messages, models, and documents separately, everything draws from one balance. See Credits explained.
Answering a visitor message (cost varies by model), processing a source into the knowledge base, voice transcription, and agent tools/skills. Internal testing in the Playground is free. Details in Credits explained.
Plan credits come with your subscription and refill each billing period; Top-up credits are bought (manually or via auto-recharge) and carry over. They’re spent in order — plan first, then top-up. See Credits explained.
Agents stop generating new answers until you top up or your plan period resets — but nothing is deleted. Turn on auto-recharge so a busy spell can’t take agents offline. See Credits explained.
Use Top Up Now for an immediate top-up, or turn on auto-recharge to top up automatically when your balance gets low. Manage your payment method through the checkout & billing portal.
Yes — that’s Pay for Access. You set a price, access duration, and message allowance, and the revenue is yours. It’s set up with help from the A2V2.ai team.
Compare and switch on the Plans page. Billing changes are Owner-only. Need capacity beyond credits? See Add-ons.

​
Team & access

Invite from the team list and assign a role. See Members & roles.
Owner can do everything, including billing and deleting the organization. Admin manages agents, knowledge, contacts, and team members. User works with agents and content without organization-level admin. Full breakdown in Members & roles.
You’ve used every team seat on your plan. Add the Team Member add-on or upgrade your plan. Seat changes are handled by the Owner.
Their invite is pending until they accept it from the email. Check the Progress column on the team list, and re-send if needed. See Members & roles.
Email is read-only on the Profile page because it’s your sign-in identity. To use a different email, contact support@a2v2.ai.

​
Data & privacy

Yes. Everything you do happens inside your organization, and your data is isolated from every other organization on the platform. See Core concepts.
Only the sources you add to its knowledge base, retrieved per question and used to ground the answer. See How answers are generated.
Yes. Delete the source from its Actions menu (⋯) and it’s removed from the agent’s index, so it no longer appears in answers. See Manage sources.
The Audit logs record activity across your organization.
Email support@a2v2.ai with your organization name and the specifics of your question.

​
Still stuck?

Troubleshooting

Symptom-by-symptom fixes for the most common issues.

Core concepts

How agents, knowledge, RAG, and credits fit together.

Quickstart

Build and embed your first agent end-to-end.

Contact support

Email us with your organization name and the agent involved.