A production SaaS platform for building AI agents from product knowledge
Echo is an internal SaaS product that lets a business shape an AI agent around its own description, instructions and knowledge, then publish it as a shareable chat or embed it on a website.
What the product needed to solve
A generic chatbot can sound confident without understanding a product. Echo needed to make agent configuration approachable for non-technical owners while keeping public answers grounded, the delivery reusable and the operating costs controlled.
- Designed a platform-owned behavioral guardrail that remains active regardless of the owner's prompt.
- Separated agent configuration, public delivery and administration into clear product contours with distinct permissions.
- Built Telegram bot deep-link authentication around short-lived, single-use login attempts and secure server sessions.
- Kept model calls behind a backend adapter so provider failures, timeouts and empty responses can be handled consistently.
The implemented product layer
Responsive marketing site, authentication flow and account workspace
Agent builder with description, instructions, knowledge and public customization
Public chat, persistent browser conversations and an embeddable website widget
Admin visibility, usage controls and production deployment on custom domains
How the system was organized
The React and TypeScript frontend talks only to a FastAPI backend. PostgreSQL stores users, agents, conversations, messages and usage data; secure HttpOnly cookies maintain sessions; Telegram confirms identity; a dedicated provider adapter handles AI requests. Frontend, API and database services are deployed separately on Railway behind production domains.
The result is a working product at echoagent.online. Users can authenticate through Telegram, create and configure an agent, share a public link, add the widget to a website and receive AI answers based on the owner's product context.
Evidence boundary
- No public traffic, revenue or conversion metrics are claimed.
- AI response quality and availability still depend on the configured external model provider.
- Paid subscriptions are outside the current production scope.
What buyers usually need to know
- What is Echo?
- Echo is a SaaS platform for creating focused AI agents from a business's own description, instructions and knowledge.
- How are agent answers kept on topic?
- The backend combines a platform-owned guardrail with the owner's configuration and knowledge before every model request.
- Can an Echo agent be embedded on another website?
- Yes. Owners can use a public agent page or add the embeddable chat widget to their own website.
- How does authentication work?
- A short-lived login attempt opens the Echo Telegram bot. After the user confirms with Start, the browser receives a secure server session without exposing a bot token to the frontend.
Need an AI SaaS product that works beyond the demo?
We design the user journey, application architecture, integrations and production delivery as one product system.