Concierge

Concierge is an in-app support SDK at conciergesdk.com, from Pixel Port LLC in St. Petersburg, Florida. It adds a native chat to iOS, Android, React Native, Expo, and Flutter apps, plus a widget for a website.

An AI agent answers first from docs you add, such as FAQs, pricing, and known issues. Replies are labeled as AI. If a user asks for a person, the thread moves to the top of the inbox and is marked high priority.

When you reply yourself, the agent pauses on that thread. The chat works for signed-out users. Call identify when someone signs in if you want their history on every device.

The SDK makes no network call when the app launches. It registers the first time the chat opens or a push token arrives. iOS uses SwiftUI, needs iOS 16 or later, and lists no extra dependencies. Android uses Kotlin and Jetpack Compose, with minSdk 24, and does not depend on Firebase.

React Native and Expo wrap those native chats. Expo Go can run a JavaScript fallback before you make a development build, installed with npx expo install @conciergesdk/react-native. The Flutter plugin concierge_sdk wraps the same native chats, needs iOS 16.4 or later and Android minSdk 24, and does not cover Flutter web.

Use the web widget there. The widget is one script tag or npm install @conciergesdk/web, with a React binding, drawn in a shadow DOM. The docs say it loads about 27 KB gzip with the page, and about 7 KB more when the chat opens, and that it makes no API call until someone opens the chat.

The API only answers browsers on websites you list under Allowed websites. Setup is one prompt from the dashboard, pasted into a coding agent, which installs the SDK with your app key. npx @conciergesdk/cli init detects a Swift, Kotlin, Expo, React Native, or Flutter project, installs the SDK, wires push, and waits for the first test message. npx @conciergesdk/cli doctor checks the install afterward. The inbox is also a remote MCP server at mcp.conciergesdk.com, with 51 tools that read and write over Streamable HTTP.

Claude Code, Cursor, and Codex can sign in with OAuth. Windsurf, VS Code with GitHub Copilot, Gemini CLI, Zed, and other clients can send a staff API key. A reply from the agent reaches the app live and by push.

The triage prompt groups open threads and drafts replies, and nothing sends until you approve. A client tool, such as Restore purchases, runs only after the user allows it. The site's Claude Code recording shows that flow on a sample app called Lumen Journal.

It lists open threads, reads a locked-purchase report, changes the restore code, replies, and flags another thread as high priority. There is no free plan and no trial. Before you pay, you can sign up, add apps, write help articles, and send test messages from up to 3 of your own devices.

Those test messages do not land in the spam folder. AI replies and messages from real users start when you subscribe. You can cancel in the billing portal, and Pro stays on until the end of the month you paid for.

Pro is one price for unlimited apps and seats, with 1,000 AI replies a month. The first 50 organizations to pay get $19 a month, or $190 a year, locked while that subscription stays active without a break. Canceling ends the lock, and a later subscription pays the price then offered to new subscribers.

Monthly and yearly share the same 50 spots, one per organization. A spot counts once the first payment succeeds and is not released if you cancel. After those 50, Pro is $49 a month, or $400 a year.

Prices are in US dollars and exclude tax. When the 1,000 replies are used, AI replies pause until the next billing month. Concierge does not charge overage unless you agree to it separately.

Scale is priced on request and adds higher AI volume, SSO, an audit log, data residency, and an SLA. Two-way translation, Slack replies, and in-app analytics are marked Soon on the pricing section. The refund desk and billing heads-ups stay off until you turn them on.

The refund desk answers Apple's consumption request inside the 12 hour window, from the refund policy you set. Review replies are drafted in the reviewer's language and do not post until you approve them. Chat buttons, labels, and the AI label ship in 34 languages, including right to left, and follow the device language.

The title and welcome line you set in the dashboard stay as you wrote them, and anything you set in code wins. Self-hosting is not a supported product. The SDKs accept a custom server URL, which Concierge uses against a local API in development, and Concierge does not ship a packaged server.

The terms, last updated November 2, 2026, name Pixel Port LLC, a Florida company in St. Petersburg. Servers and the database are in Canada, on OVHcloud in Beauharnois, behind Cloudflare.

Attachments are in private Cloudflare R2 storage in North America. Concierge's October 2026 eval puts the average model cost of one AI reply at $0.00075, with tool calls included. That figure is Concierge's measurement.

Pricing: Checked November 8, 2026 on https://conciergesdk.com/#pricing and https://conciergesdk.com/legal/terms/ (terms last updated November 2, 2026). No free plan and no trial. Test messages from up to 3 of your own devices before you subscribe. Pro is $19 a month or $190 a year for the first 50 organizations, locked while the subscription stays active. After that, $49 a month or $400 a year. Unlimited apps and seats, 1,000 AI replies a month, then AI replies pause until the next billing month. No overage unless you agree to one. Prices are in US dollars and exclude tax. Scale is priced on request. /pricing returns 404. The prices are on the homepage pricing section.