Add a Masheev assistant to your product
An embed SDK, a typed API client and custom tools, so your product gets an assistant that answers and acts without you building a chat stack. Developer access is in beta.
Card required. No charge if you cancel within 14 days. Available to businesses in the UK and US.
- Published on npm
- Typed end to end
- React, JavaScript and headless
import { useMasheev } from "@masheev/embed-sdk/react";
function App() {
const masheev = useMasheev({
inboxId: "inb_123",
mode: "prompt-input", // or "chat-widget" / "embedded"
});
return <button onClick={() => masheev.open()}>Ask AI</button>;
}What is Masheev for developers?
Masheev gives developers the pieces to put an AI assistant inside their own product: an embed SDK (React hook, vanilla JS or headless WebSocket), a type-safe API client, custom tools the assistant can call in your app, and an MCP server for supported workspace operations.
Developer access is in beta. It covers the supported API and widget capabilities; provider-gated channels remain private betas, and durable custom webhooks are not available yet. Sessions are protected by signed short-lived tokens, bot protection and rate limits, with data held in the EU or US region.
Works with the assistants you already use
Masheev publishes an MCP server and open Agent Skills. Coding agents and assistants such as Claude, Codex, Cursor, Copilot and Gemini can learn the SDK and work in your workspace.
- Claude
- Codex
- Cursor
- Copilot
- Gemini
Connect an assistant over MCP
npx @masheev/mcpTeach your coding agent the SDK
npx skills add masheev/skillsThe building blocks
Embed SDK for React, JavaScript and headless
Add a chat widget, a prompt bar or an embedded panel with the React hook or a script tag. Or go headless over WebSocket and render your own UI.
Custom tools the AI can call
Define typed tools with Zod or JSON Schema. The assistant calls them in your app to look up data, take actions or show a card, with optional user approval.
Type-safe API client
@masheev/client gives you typed access to inboxes, conversations, contacts, AI agents and knowledge, from React, a server or React Native.
MCP server
Connect Claude Desktop, Cursor or your own agent to your workspace over the Model Context Protocol. The assistant can discover and call the Masheev API with your token's permissions.
Three ways to integrate
Use the prebuilt interface to get going, or go headless for full control. The platform underneath is the same.
The React hook and script tag render a finished assistant in one of three modes: prompt-input (a prompt bar), chat-widget or embedded. For your own design, the headless entry point streams messages over WebSocket and you render every element.
// 1. React hook
import { useMasheev } from "@masheev/embed-sdk/react";
// 2. Vanilla JS or script tag
import { init, open } from "@masheev/embed-sdk";
init({ inboxId: "inb_123", mode: "chat-widget" });
// 3. Headless: your own UI, streamed over WebSocket
import { MasheevProvider, useMasheev } from "@masheev/embed-sdk/headless";Frequently asked questions
Explore more
Build with Masheev
Create a workspace, copy your inbox ID and add the assistant to your product. Developer access is in beta.
Card required. No charge if you cancel within 14 days. Available to businesses in the UK and US.