NativeX vs SofaCode

NativeX is a visual playground for Apple's on-device intelligence — build prompt pipelines as a graph of cards, wire them together with variables, and run them right on your iPhone. Built for iOS engineers, prompt engineers, and the AI-curious: if you want to know what Apple's on-device model can actually do, this is the fastest way to find out. FREE AND PRIVATE Every run executes on the Apple foundation model that ships with your device. No account, no API key, no server: prompts, photos, and recordings never leave your iPhone, and inference costs nothing. LEARN THE REAL API Every card mirrors one Apple framework call and names the exact Swift symbol — LanguageModelSession, NLTokenizer, RecognizeTextRequest, SpeechTranscriber, etc. Configure it in the inspector, run it, read the output. What you learn transfers line-for-line to Xcode. BUILD PROMPT PIPELINES • Prompt groups assemble instructions, few-shot examples, history, and the curreturn into a single transcript • Guided generation: constrain output to JSON with registered schemas, or design your own schema on the fly — no rebuild • 17 native ops: tokenize, detect language, named entities, sentiment, OCR a photo, read a barcode, transcribe speech, speak text aloud, regex, JSON extraction, and more. Wait for more! • Human-in-the-loop pauses: type or dictate an answer mid-run EVALUATE LIKE A PRO — LLM OPS, ON-DEVICE NativeX brings the modern LLM-ops workflow — observability, datasets, evaluations — to the model in your pocket: • Run History logs every execution as a step-level trace — assembled prompt, output, timing, token estimates • Datasets batch-run a graph over a table of records, sequential or parallel • LLM-as-judge scoring runs on-device too • Re-run, reproduce, and compare past runs REQUIREMENTS NativeX runs the Apple Intelligence on-device model. That requires an iPhone 15 Pro or newer (A17 Pro chip or later) on iOS 26 or later, with Apple Intelligence enabled in Settings.

SofaCode turns your phone or tablet into a remote control for the AI coding assistant already running on your computer — Claude Code, Codex, GitHub Copilot Chat or Local LLMs in VS Code. Vibe coding, off the desk. Talk to your agent from the couch. Approve its actions from the kitchen. Watch it work from anywhere on your network. NOT JUST VS CODE — NOW IN BETA SofaCode started as a VS Code remote. Now it drives your agent in the VS Code editor you actually use: Cursor, VS Code Insiders, VSCodium, Devin (formerly Windsurf) and Antigravity (Google's IDE) — all in beta, right alongside stock VS Code. Same deck, same voice, same real approvals, whichever one you open. (Copilot Chat features apply only where Copilot Chat itself runs: VS Code and Insiders.) HOW IT WORKS A small companion daemon runs on your computer next to VS Code — free for Mac, Windows and Linux at https://sofacode.app. The app connects to it directly over your own Wi-Fi, VPN, or mesh network. Nothing routes through a cloud: no relay servers, no accounts, and your code, voice, and sessions never leave your network. WHAT YOU CAN DO • Push-to-talk dictation straight into your coding agent • Live deck that shows working / needs-you / error states at a glance • Approve or deny permission prompts remotely — real approvals, not keystroke hacks • Full chat with code boxes, task checklists, and session transcripts • Drive all three assistants alike — skills, model picker, usage, tasks, permissions and more work the same for Claude Code, GitHub Copilot Chat and OpenAI Codex • Browse sessions across machines, switch windows, open projects • File browser, terminal, saved prompts, themes (including LCARS), TTS read-aloud • Works across editors — VS Code, plus Cursor, Insiders, VSCodium, Devin and Antigravity in beta • Multiple machines and multiple devices at once WHAT YOU NEED A computer running VS Code with Claude Code, Codex or GitHub Copilot Chat — your own assistant subscription — plus the free SofaCode daemon. Copilot Chat can also run local models from LM Studio or Ollama. Setup takes a few minutes with QR pairing. SofaCode is an independent product and is not affiliated with OpenAI, Anthropic, GitHub, or Microsoft. PRICING One subscription, one week free, then $5/month or $50/year. Cancel anytime in Settings. Terms of Use (EULA): https://www.apple.com/legal/internet-services/itunes/dev/stdeula/ Privacy Policy: https://sofacode.codedatda.casa/privacy.html Your coding agent. In your comfy place. by Codedatda.casa