Getting Started
Amba is the agentic backend-as-a-service for utility mobile apps — built for vibe coders and AI agents.
What is Amba?
Amba is an Agentic Mobile Backend-as-a-Service. It gives utility apps (affirmations, calorie counters, prayer apps, lock-screen apps) the domain-specific primitives they always end up rebuilding — user auth, push campaigns, segments, remote config, content libraries, streaks, XP, entitlements — with a single SDK and a single API.
The twist: it is designed from the ground up to be managed by an AI coding agent via MCP. Instead of clicking around a dashboard to create a segment and a push campaign, you tell Cursor or Claude Code what you want and the agent does it.
What you get
Out of the box, every Amba project gives your app:
- User auth — anonymous, Apple, Google, email / password, with account linking that preserves anonymous history.
- Push notifications — APNs + FCM, campaigns, segment targeting, scheduled delivery.
- User segments — rule-based, 12 operators, re-evaluated every 15 minutes.
- Remote config — typed key / value pairs with segment-based and percentage-rollout overrides.
- Content libraries — scheduled daily tips, rotating quotes, sequential lessons.
- Streaks — daily / weekly periods, grace windows, freezes.
- XP and levels — automatic XP awards driven by event rules.
- Achievements and badges — auto-unlock based on progress criteria.
- Leaderboards — XP, streaks, or any custom metric.
- Challenges — time-limited goals with rewards.
- Friends, groups, feeds, messaging — a full social layer.
- Virtual currencies, stores, inventory — in-app economy.
- Entitlements — RevenueCat + Superwall integrations via verified webhooks.
Who is this for?
Vibe coders. Developers who ship fast with AI tools (Cursor, Claude Code, Copilot). You want a backend that Just Works without reading 47 pages of Firebase docs.
AI agents. Amba ships an MCP server with a large tool surface. Your agent creates push campaigns, defines segments, sets up streaks, configures XP rules, and manages content libraries directly — you write app code, not backend glue.
How does it compare?
| Feature | Firebase | Supabase | PlayFab | Amba |
|---|---|---|---|---|
| Auth | Yes | Yes | Yes | Yes |
| Push notifications | FCM only | No | Yes | Yes (APNs + FCM) |
| User segments | No | No | Yes | Yes (12 operators) |
| Remote config | Yes | No | Yes | Yes (segment-targeted) |
| Content scheduling | No | No | No | Yes |
| Streaks | No | No | No | Yes |
| XP / levels | No | No | Yes | Yes |
| Achievements | No | No | Yes | Yes |
| Leaderboards | No | No | Yes | Yes |
| Virtual economy | No | No | Yes | Yes |
| Social (friends, groups, feeds) | No | No | Yes | Yes |
| MCP server for AI agents | No | No | No | Yes |
| RevenueCat / Superwall integration | No | No | No | Yes |
How you use Amba
Three surfaces:
- CLI (
npx amba init) — bootstrap a project, mint API keys, drop agent context files. - MCP server (stdio) — let your AI agent (Cursor, Claude Code, Zed) call Amba directly as typed tools.
@amba/clientSDK +@amba/expowrapper — what your app code talks to.
Each project is isolated: your data lives in its own dedicated database. No shared tables, no cross-project leakage.
Next steps
- Quickstart — five-minute Expo setup.
- Kitchen-sink Expo tutorial — step-by-step to-do app that exercises every SDK module.
- SDK reference — explore the client SDK.
- MCP server — set up AI-agent access.