comparison

AgentKit Cloud vs Firebase AI Logic

Firebase AI Logic locks you to Google's models and attests the app, not the user. AgentKit Cloud routes Anthropic, OpenAI, and Gemini, verifies the end user, and caps spend per user. Your keys stay yours, and the same SDK calls any provider directly if you leave.

Firebase AI Logic is Google's mobile SDK for calling Gemini, with App Check attestation (as of June 2026).

AgentKit CloudFirebase AI Logic
Model providersAnthropic, OpenAI, Gemini, and Apple on-deviceGoogle's models only (Gemini)
Provider keys off the deviceEncrypted server-side; the app ships a publishable keyHeld by Google's backend
End-user identity verified per requestA JWT from your trusted issuer, checked every requestApp Check attests the app, not the user
Hard spend cap per userA hard ceiling per user and per project. Hit it, requests stop.Project limits and a per-user rate throttle, no per-user spend cap
Subscription-tier model routingBind tiers to RevenueCat entitlementsNot native
Change models without an app updateName a tier; remap it server-sideClient, or Remote Config
On-device optionApple Foundation Models on-device, cloud when neededCloud only on iOS
In-app agent loop with tools, guards, undoRuns in your app with guards, undo, and run limitsFunction calling, no guard or undo layer
Device attestationOptional App Attest binds requests to real devicesApp Check with App Attest or DeviceCheck
Backend to build and maintainNone. AgentKit Cloud is the backend.Google runs the relay
  • More than Google's models

    Route the same app to Anthropic, OpenAI, or Gemini, and move a tier between them server-side.

  • Identity, not just attestation

    App Check proves the app is genuine. AgentKit also verifies who the end user is, with a JWT from Apple, Firebase, or your own issuer, on every request.

  • Per-user billing and quotas

    Attribute spend per end user, cap it per user and per project, and bind tiers to RevenueCat entitlements.

Skip the server. Ship the agent.