Private beta for teams shipping AI in client apps

Ship client AI with a control plane you can actually debug

Keep provider keys off clients, govern every AI request, and debug production issues without spelunking raw JSON. Wavelike routes hosted, self-hosted, and on-device model calls through one observable gateway.

For teams shipping AI in native and web apps. We'll follow up with next steps.

Payloads, schemas + latency
Production visibility
Hosted, BYOK + tunnels
Model routing
Auth, quotas + attribution
Per-user controls
import Wavelike Wavelike.set(environment: .production)Wavelike.set(appId: "app_live_...")Wavelike.set(auth: .backend(userId: user.id, key: userKey)) let model = Wavelike.model(  for: ModelIdentifier(ModelClient.self, id: "support-chat")) let receipt = try await Wavelike.uploadFile(at: receiptURL) let response = try await model.sendText(history: [  Message(role: .user, parts: [    .inputText("Summarize this receipt"),    .inputFile(uploadedFile: receipt),  ])])

Responses

Streaming, tools + schemas

Attribution

BYOK, keys + app users

Forensics

Files, media + local SDKs

Routes client traffic to the providers, private endpoints, and API surfaces your apps already use

OpenAI
Anthropic
OpenRouter
Together AI
Google
AWS Bedrock
xAI
Fireworks AI
Custom endpoints
Self-hosted tunnels

Built for developers

One operating layer for the AI your app ships.

Wavelike brings identity, routing, tunnels, quotas, files, attribution, traces, and visual diagnostics into the path between your users and the models they call.

Client AI Gateway

Expose one OpenAI-compatible API for client AI while Wavelike handles credentials, request validation, model routing, stack configuration, and response normalization.

BYOK Routing

Route stacks through Wavelike-managed providers, organization-owned keys, custom model IDs, hosted endpoints, or self-hosted tunnels without putting upstream secrets in client apps.

Identity

Gate requests with dashboard-managed auth strategies: API keys, backend-issued user keys, Connect, Enterprise SSO, Apple DeviceCheck, or Google Play Integrity.

Spend Intelligence

Track cost, tokens, requests, trends, audio usage, BYOK source, and composition by app, user, API key, stack, provider, and model so product teams can see what is driving spend.

Request Observability

Debug gateway and on-device calls with searchable logs, API-key and user attribution, JSONL export, latency phases, provider errors, stored payloads, output schemas, thinking blocks, and visual payload views.

Quota Policies

Layer policies for cost, tokens, requests, or weighted request units across all users, auth strategies, API keys, individual users, or model stacks, with estimated reservations before provider calls.

Model Stacks

Configure model stacks, pinned versions, system behavior, compaction, image, and audio slots from the dashboard, then route by provider priority, BYOK selection, or private custom provider.

SDKs

Use the iOS SDK, WavelikeUI chat components, on-device observability modes, file upload helpers, OpenAI-compatible HTTP endpoints, and protected docs with API reference content for implementation teams.

Every request passes through policy

Requests pass through strategy validation, app-user resolution, Connect or Enterprise SSO entitlement checks, model resolution, quota checks, canonical model routing, provider or BYOK selection, private tunnel dispatch, error normalization, canonical payload capture, async response logging, SDK metadata capture, API-key attribution, and usage recording before the response returns to the client.

Integrate without a backend rewrite, or a backend at all

Use Wavelike behind an existing service, or let Wavelike be the backend your client calls for identity, model access, file uploads, inference, provider credentials, BYOK models, self-hosted model tunnels, quotas, on-device observability, and exportable usage records.

Turn production requests into evidence

The request detail page turns canonical payload JSON into a readable visual debugger for prompts, responses, stream events, tool calls, tool results, output schemas, thinking blocks, uploaded-file references, image inputs and outputs, audio, size metadata, SHA fingerprints, SDK runtime metadata, local-model records, and provider or pipeline failures. Filter by caller, model, stack, media, metadata, status, or latency, then export the matching request history as JSONL.

Private beta access

Join the private beta for controlled AI in client apps.

Get early access to the dashboard, API gateway, iOS SDK, auth strategies, Connect and Enterprise SSO, file uploads, quota policies, BYOK providers, self-hosted custom providers, on-device observability, dashboard-managed model stacks, visual request forensics, JSONL export, and per-user/API-key spend analytics.

Join the waitlist and we'll follow up with private beta access.