Your Echo, inside every AI app you already use

Overview

The Model Context Protocol is the emerging standard for letting AI apps use external tools, and Echo AI is a first-class remote MCP server. One HTTPS endpoint exposes eight tools, four resources and four prompts, all behind OAuth 2.1 with PKCE and Dynamic Client Registration, so any compliant client can register itself and ask you for consent. Once connected, an assistant can list the Echos your account can access, read their configuration and analytics, browse recent conversations and message history, and send messages to an Echo on your behalf. It also covers the build path: get_embed_snippet returns the exact widget script for your site, and get_headless_integration returns the REST base URL, every endpoint (streaming chat, config, catalog, checkout, booking, tickets, voice transcribe and speak, session restore) and a ready-to-paste React component, so a coding assistant can wire an Echo into your product in one pass while the knowledge, personality and integrations stay managed in the Echo AI editor. Read tools are free; only chat calls draw on your normal credit balance, and plan rate limits still apply. You can revoke a client at any time. Connecting external AI apps through MCP requires a Business plan or higher.

What you get

• One remote URL - Add auth.echoai.so/functions/v1/mcp as a remote MCP server. No local install, no SDK, no proxy to host.
• OAuth 2.1 with PKCE - Clients register themselves through Dynamic Client Registration and you approve access on an Echo AI consent screen. Revoke any client at any time.
• Eight tools - List and read Echos, pull the embed snippet, read analytics, browse conversations and message history, chat with an Echo, and get the full headless integration guide.
• Ship the integration - get_embed_snippet returns the widget script for your site, and get_headless_integration returns every REST endpoint plus a ready-to-paste React component.
• Resources - echo://echos, echo://echo/{id}, echo://echo/{id}/analytics and echo://docs/api are readable as MCP resources, so assistants can pull context without a tool call.
• Built-in prompts - audit_echo, summarize_echo_activity, embed_echo_on_site and test_echo_answer show up as slash commands in Claude, Cursor and VS Code.

How it works

• 1. Copy the server URL - Grab auth.echoai.so/functions/v1/mcp from the docs at echoai.so/api.
• 2. Add it to your AI app - Paste it into ChatGPT, Claude, Cursor, Codex, Windsurf, Replit or Lovable as a remote MCP server.
• 3. Approve the OAuth screen - Sign in to Echo AI, approve the client, and your assistant can use every Echo your account can access.

Who it is for

• Put your Echo in your app or website - Ask your AI coding tool to embed your Echo and it pulls the real snippet from your account: get_embed_snippet for the hosted widget, get_headless_integration for a fully custom chat UI with the streaming chat, voice, catalog, checkout and booking endpoints plus a working React component. No copying keys or docs by hand.
• Build and audit Echos from your IDE - In Cursor, Codex, Windsurf or Claude Code, run audit_echo to review a persona, FAQs and preset questions, check analytics, and read last week's conversations without opening the editor.
• Chat with your Echo from ChatGPT - Test answers before customers see them, replay a conversation, and send messages to any of your Echos without leaving the assistant.
• Report on Echo performance - Ask for conversation, message and visitor counts across your Echos, then have the assistant write the weekly summary or drop it into the doc you already work in.

Keep exploring

Frequently asked questions