Engine Gateway
Central entry point of the engine platform. It sits between SDK consumers and engine instances and hides each engine's language and runtime behind one uniform API.
Search every APPNEURAL gateway by traffic family, lifecycle, and runtime—from the edge API gateway and the Envoy control plane to the engine and AI gateways, plus the gateway types still waiting for a repository.
20tracked gateways
0in production
9repositories · 11 planned types
0high priority (P0–P1)
Central entry point of the engine platform. It sits between SDK consumers and engine instances and hides each engine's language and runtime behind one uniform API.
Control plane for API routing intent: immutable configurations, approval, publication, and history, while Envoy serves application traffic independently.
One hostname in front of 27 platform Workers at api.appneural.com/{service}/{path}. The prefix is stripped at the edge, every hop is a Service Binding, and it serves its own OpenAPI reference.
Filter the directory by traffic family and lifecycle.
Controls outbound calls from services with allow-lists, SSRF protection, secret injection, and an audit trail of third-party requests.
Identity-aware proxy that checks sessions and tokens before traffic reaches internal apps and admin panels.
Partner-facing API surface with per-partner keys, quotas, contract versions, and usage reports.
MQTT and HTTP entry point for devices, covering device identity, telemetry ingestion, and command delivery.
Accepts events from external producers, checks schema and signature, and hands them to queues for platform consumers.
Provider-agnostic SMS and OTP delivery with sender routing, templates, retries, and delivery receipts.
One checkout and payment API in front of payment providers for cards, UPI, and wallets, with idempotent charges, refunds, and signed provider webhooks.
WebSocket and server-sent-event entry point that authenticates long-lived connections and fans messages out to subscribers.
Federated GraphQL endpoint that composes service schemas into one graph, with query cost limits and persisted queries.
One authenticated entry point for agents to reach APPNEURAL MCP servers, with per-tool authorization, audit logs, and rate limits.
Multi-provider model routing with per-product budgets, response caching, fallbacks, and usage metering, taking over the role of the legacy AI Gateway.
Fronted Microsoft Foundry for every product through one POST /v1/infer endpoint, with service-token auth, request validation, PII redaction, and a KV-backed daily budget per product.
Earlier engine-layer gateway for API routing, authentication, rate limits, quotas, versioning, the API catalog, and consumer access. Kept in the old engines organization.
Placeholder for the DatePe client app's API gateway. The repository was created in April 2025 and is still empty.
Planned architecture and developer OS for the gateway layer: routing, authentication, throttling, request policies, and aggregation. The repository is still empty.
Planned admin panel for traffic routing and API gateway policies: upstreams, routes, service discovery, throttling, traffic metrics, and configuration rollback. Documentation only so far.
Single API hostname, api.erp.appneural.com, that dispatches requests to all 30 MicroERP Workers.
Central entry point of the engine platform. It sits between SDK consumers and engine instances and hides each engine's language and runtime behind one uniform API.
Control plane for API routing intent: immutable configurations, approval, publication, and history, while Envoy serves application traffic independently.
One hostname in front of 27 platform Workers at api.appneural.com/{service}/{path}. The prefix is stripped at the edge, every hop is a Service Binding, and it serves its own OpenAPI reference.
Record each gateway's traffic family, runtime, boundary, and repository. Planned gateway types stay marked as Reference until a repository exists.