- Add Gateway Pages with static-site projects, immutable deployments, mutable tags, runtime configuration, deploy tokens, manual archive uploads, node placement, migration, retention, and wildcard previews - Add managed path-based Additional Routes for addresses, Docker containers, deployments, and Pages tags, including per-route WebSocket, buffering, timeout, prefix stripping, and Secure Link controls - Add scoped Pages and ingress automation to AI Workspace, MCP, search, audit, notifications, and resource navigation - Enforce plan entitlements consistently across backend APIs and frontend surfaces, including the Personal requirement for Pages and fail-closed feature access - Replace the built-in inference engine with a privately managed OpenCodex core that Gateway can install, verify, update, repair, and automatically roll back - Add signed inference-core release discovery, download progress, health and version reporting, installer integration, and lifecycle recovery after interrupted operations - Keep provider selection, failover, quotas, limits, accounting, pricing, and access control in Gateway while delegating protocol execution and compaction to OpenCodex - Add priority, even-distribution, and quota-balanced routing across eligible provider accounts with affinity, reserve thresholds, cooldowns, and pre-output failover - Add managed ChatGPT, OpenAI API, Claude, Anthropic API, Kimi, Moonshot, xAI, OpenRouter, opencode go, Z.AI GLM Coding Plan, Alibaba Token Plan, and custom OpenAI-compatible provider connections - Add device authorization for ChatGPT and Kimi, subscription quota synchronization, live provider model discovery, reasoning metadata, context overrides, and account-specific availability - Add logical inference models with stable public IDs, model capabilities, reasoning mappings, access rules, API pricing, subscription multipliers, and configurable context and compaction limits - Expose stable OpenAI and Anthropic inference APIs for Models, Responses, Chat Completions, Messages, images, search, realtime calls, streaming, continuation, and remote compaction - Add protected inference accounting with atomic reservations, immutable settlement records, per-user API and subscription budgets, recovery reserves, and ChatGPT Fast-tier accounting - Add the `@wiolett/gateway-inference` interactive companion for secure Codex and Claude Code setup, catalog synchronization, diagnostics, repair, removal, and guarded logout - Preserve existing Codex and Claude Code configuration during setup and removal while storing dedicated inference credentials outside harness configuration files - Harden public health endpoints, protected routes, Docker emergency actions, archive validation, route retargeting, and inference cancellation and failure handling - Improve Gateway Inference setup, provider, model, endpoint, quota, update, and lifecycle UI across the browser installer, settings, profile, and AI Workspace - Require existing installations to install the managed inference core after upgrading and reconnect supported harnesses when the endpoint contract changes