Yuva

io.github.productdevbookv0.0.4Updated Oct 9, 2026

Customer support inbox: search, read, triage and draft replies to e-mail and chat.

VerifiedStreamable HTTPWeb executableBusiness & CommerceCommunication & Collaboration

Overview

AI-generated overview

Lets an assistant search, read, triage and draft replies to customer support e-mail and chat in a Yuva inbox.

What it does
Connects an assistant to a Yuva customer messaging inbox so it can search and read support e-mail and live chat conversations, triage them, and write draft replies for a team member to send. Yuva itself runs no model; the assistant works signed in as a user, and scoped API keys and inbox membership limit what it can reach. The manifest does not list individual tool names.
When to use it
Useful when a team runs customer support in Yuva and wants an assistant to help sort incoming conversations and prepare reply drafts. It is aimed at self-hosted Yuva deployments rather than a general-purpose inbox.
Requirements
A Yuva server URL (YUVA_URL) and a Yuva API key (YUVA_API_KEY); the key's scopes and inboxes limit what the assistant can do. Can run as a remote endpoint or as a local process, and the server itself is self-hosted (one Go binary with Postgres, or a Docker image).
Before you install
The project is described as pre-alpha: the API and database schema can change without migration paths and the code has not had an independent security audit, so it should be tried with test data rather than real customer conversations. The API key (YUVA_API_KEY) is a secret, and the assistant can read customer conversations and write drafts.

Installation

In SourceWeft

  1. Open Yuva in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Web executable via Streamable HTTP. Remote servers run from the web runtime once configured in a workspace.

Other MCP clients

Add this to your client's mcpServers config.

{
  "mcpServers": {
    "yuva": {
      "type": "http",
      "url": "https://{yuva_host}/mcp"
    }
  }
}

README

[Yuva logo]

Yuva

One inbox for every product you run.

Open-source customer messaging: support e-mail, live chat and in-app conversations from all of your products in one shared inbox. Self-hosted, one Go binary with Postgres.

Website · Docs · API · Releases · Roadmap

[Release] [CI] [License: AGPL-3.0] [SDKs: MIT]

[The Yuva panel: conversations from three products in one list, an in-app conversation with a note and a reply, and the contact's details.]

[!WARNING] Pre-alpha. The API and database schema can change without migration paths, and the code has not had an independent security audit. Try it with test data, not real customer conversations.

Why Yuva

  • Every product, one inbox. Each product is an inbox with its own branding, language, hours and team. Members see only the inboxes they were given.
  • Every channel, one conversation. E-mail with proper threading, live chat with the <yuva-chat> web component, in-app messages and feedback through native iOS and Android SDKs, and an API for everything else.
  • Your users stay yours. Your backend signs a short-lived identity token; Yuva never sees your user database. Signed webhooks let your backend send its own push notifications.
  • Headless when you want it. Scoped API keys, bots that write drafts for a member to send, an event feed and a headless JS client let you build your own inbox, chat and automations.
  • Works with AI assistants. A built-in MCP server lets Claude, ChatGPT, Cursor and other assistants triage and draft replies, signed in as you with OAuth. Yuva runs no model itself.
  • Small to run. One Docker image and Postgres. No Redis, no separate workers.

Quick start

sh
docker pull ghcr.io/productdevbook/yuva:0.0.4
  1. Write compose.yaml and .env from the install guide.
  2. Start it: docker compose up -d.
  3. Create the first owner: docker compose exec yuva /yuva bootstrap --email [email protected] --workspace "Example".
  4. Put it behind HTTPS and sign in with the code that arrives by e-mail.

Documentation

Get startedChannelsIntegrateBuild on it
InstallE-mailIdentity tokensHeadless
ConfigurationWeb widgetWebhooksAI assistants (MCP)
OperationsMobile SDKsAPI contractExamples

How it is built: architecture. What is next: roadmap. What changed: changelog.

Repository

PathWhatLicense
api/Go server: API, realtime, e-mail, jobsAGPL-3.0
web/Team panel (React)AGPL-3.0
edge/Cloudflare Email Worker for inbound mailAGPL-3.0
site/Website and documentationAGPL-3.0
deploy/Docker and ComposeAGPL-3.0
openapi/API contractMIT
sdk/jsuseyuva: headless client, <yuva-chat>, React hooks, typed /v1 clientMIT
sdk/swift, sdk/kotliniOS and Android SDKsMIT
sdk/goTyped /v1 client, identity tokens and webhook verificationMIT
examples/Headless chat, event feed and draft botMIT

Contributing

Issues and pull requests are welcome; read CONTRIBUTING.md first. Contributions are made under the CLA. Report security issues privately as described in SECURITY.md.

License

The server and panel are AGPL-3.0; the SDKs and the API contract are MIT, so they ship inside closed-source apps without copyleft obligations. Details in LICENSING.md. The name and logo are covered by TRADEMARK.md.

“Yuva” is Turkish for nest, or home.

Source: README.md at commit 010591b

Tools

0
Tool metadata has not been indexed yet.

Version history

1
  1. v0.0.4LatestOct 9, 2026