Why we built it?
Cloud‑only AI can mean unpredictable spend, tail latency, and privacy headaches. Consumer hardware (laptops with NPUs/GPUs; small edge boxes) is now good enough for many real tasks. We want agents that get work done, keep sensitive work local by default, and are easy to govern.
What it is (today) - A desktop/runtime that can run models locally and optionally connect to remote backends you allow - Models (30 providers): OpenAI, Anthropic, Google, Azure, NVIDIA NIM, Hugging Face, Together, Mistral, Perplexity, OpenRouter, Cohere, DeepSeek, plus local engines like Ollama, LM Studio, vLLM, KoboldCPP, RealTimeX - Native Search (opt‑in): Google, DuckDuckGo, Bing, Startpage, Tavily, SearXNG, Brave Search - RAG sources: PDF, Word, Excel, CSV, Markdown, and websites - MCP tools/servers: Remote and local (Model Context Protocol) - An agent workbench to design tool‑using agents with tracing/evals
What’s different - Local‑first to minimize cost (use hardware you already own) - You choose the backend per agent or per step (local by default; cloud when you need powerful LLMs) - Data stays put when you need it to (in‑device or in‑region) - Agents, not just chat: build flows that run tasks end‑to‑end and post results
How it works (short) - Install RealTimeX.ai from the website https://realtimex.ai. - Connect what you allow: choose the model provider (local + remote), enable optional search tools, add RAG sources (file/web). You can attach MCP tools/servers to extend functionality. - Choose where each step runs: set a default backend (usually local); heavy steps will automatically switch to your selected remote provider. - Create and run agents: Drag & Drop or chat with the Assistant to build an “agentic flow.” Create inside the app → run inside the app.
What feedback would help most - Backend selection UX (defaults, per‑step overrides) - Agent ergonomics (drag‑drop vs. chat‑to‑build) - Observability: which traces/metrics make you trust agents running real work?
Who we are / how to reach us I’m Trung Le (founder). Happy to answer anything here.