AI Agents Platform — Build Customer-Facing AI Agents
The AI Agents Platform by Zenovae is a workbench for building, grounding, and deploying customer-facing AI agents. Upload your documents, get cited answers, and ship an embeddable chat widget or hosted link in under 30 minutes.
What It Does
The platform is not a chatbot builder, a vector database, or a model host. It is a complete workbench where you ship an AI agent — customers ask questions, answers come from your documents with citations, and you deploy via embeddable widget, hosted URL, or internal tool behind your login.
Key Features
- 5 Agent Templates: Customer Support, Sales, Internal Knowledge Base, Intake/Triage, and Custom — each with a tested system prompt and suggested knowledge sources.
- Knowledge Grounding with Citations: Upload PDFs or point at URLs. The platform handles chunking, embedding, and retrieval (RAG). Every answer cites the source document chunk. Click a citation to see the original text. If the answer is not in the documents, the agent says so instead of hallucinating.
- Draft/Live Split: Edits stay in draft until you click Publish. Live agents never change until you say so. Rollback is one click.
- Role-Based Access: Owner edits, Admin publishes, Member reads, Viewer watches. Enforced server-side.
- Usage Billing: Token costs pass through with no markup. Per-conversation invoices. Plan limits are warnings, not hard stops. For example, GPT-4 costs $0.0024 per 1K tokens — you pay exactly that, with zero markup.
- Bring Your Own Model: OpenAI, Anthropic, or any OpenAI-compatible provider. Bring your own API key. Switch models at any time.
- Three Deployment Options: Embed a chat widget on your site with one line of HTML, share a hosted public URL with no code, or keep it behind your login for internal use.
- Integrations: Webhooks out, REST API in, bring your own model, SSO protection (Google, SAML — coming Q3).
How It Works — 5 Steps
- Pick a Template: Choose from customer support, sales, internal KB, intake, or custom.
- Drop in Docs: Upload PDFs or point at URLs. Chunking and embedding happen automatically on upload. No pipeline to wire.
- Test in Preview: A sandbox conversation with citations visible and tool calls logged. Drafts never touch production.
- Publish When Ready: Draft and live are separate. You pick the moment. Rollback is one click.
- Ship It Three Ways: Embed a widget on your site, share a public URL, or keep it behind your login for internal use.
Use Cases
Customer Support Widget
Drop the chat on your site. Cite your refund policy. Configurable fallback message when the agent doesn't know the answer. Grounded in your own documents.
Internal Knowledge Agent
Behind your SSO. Engineers ask questions, documentation answers, citations included. Role-based access enforced server-side.
Public Docs Chat
A hosted URL you can share. Visitors ask questions, public docs answer. One line of HTML to embed.
What It Is Not
- Not a chatbot builder: You ship an agent. Customers ask questions. Answers come from your docs, with citations.
- Not a vector database: You upload PDFs or point at URLs. Chunking, embedding, and retrieval are handled. You see citations, not similarity scores.
- Not a model host: Bring your own provider. Switch between them. Costs pass through. No markup on tokens.
Frequently Asked Questions
What is the AI Agents Platform?
The AI Agents Platform is a workbench for building and deploying customer-facing AI agents. You pick from 5 templates, upload PDFs or URLs for knowledge grounding, test in a preview sandbox, and publish to a hosted URL or embed a chat widget. Most teams ship their first agent in under 30 minutes.
How does knowledge grounding work?
The platform uses retrieval-augmented generation (RAG). When you upload documents, the system automatically chunks them, generates embeddings, and stores them. When a user asks a question, the agent retrieves the most relevant document chunks and generates an answer citing the exact source. If the answer is not in the documents, the agent says so instead of hallucinating.
Can I embed the AI agent on my website?
Yes. You can ship an AI agent three ways: embed a chat widget with one line of HTML, share a hosted public URL, or keep it behind your login for internal use. All three use the same agent configuration and citation system.
Which AI models are supported?
The platform supports any OpenAI-compatible provider. You bring your own API key — OpenAI, Anthropic, and others are all supported. Token costs pass through with no markup.
How much does it cost?
During the private beta, access is by invitation. Token costs pass through with no markup — you pay exactly what the model provider charges. Plan limits are warnings, not hard stops.
Request Beta Access
The platform is in private beta. We onboard teams in small cohorts. Visit the waitlist page to request access — we'll email you when a slot opens, usually within 1–2 weeks.
Zenovae — Build and ship customer-facing AI agents with cited, grounded answers.