DeliteAI

Overview

DeliteAI is the world's first open-source, fully on-device agentic AI platform built for mobile devices. With DeliteAI, developers can build, deploy and scale AI-native experiences directly on smartphones, with zero reliance on cloud infrastructure. This full-stack platform includes a production-ready SDK with an optimized on-device inference engine and the industry's first Python runtime for orchestrating agentic workflows on mobile devices. DeliteAI brings LLMs, multimodal models and AI agents directly onto user devices—no cloud roundtrips, no GPU farms on Cloud. Just blazing-fast, privacy-first intelligence running locally, all built on open standards. The platform features context engineering, structured session memory, tool-calling capabilities, and multi-step agentic workflows—all running entirely on-device. DeliteAI integrates seamlessly with industry-standard runtimes like ONNX and ExecuTorch, abstracting away diverse mobile hardware to enable high-performance AI without data ever leaving the user's device.

Key Highlights

  • Open-source, fully on-device agentic AI platform
  • Python runtime for orchestrating agentic workflows
  • Tool-calling and multi-step agentic workflows
  • Zero cloud dependency, complete privacy

Features

DeliteAI SDK

Production-ready SDK with an optimized on-device inference engine and the industry's first Python runtime for orchestrating agentic workflows on mobile devices using Python scripts. Workflows and models can be dynamically updated on-the-fly via a pluggable SaaS backend without requiring new app releases.

Agent Marketplace

A curated library of prebuilt, plug-and-play AI agents for tasks like summarization, recommendations, speech processing, and more. Developers can build and publish agents via the marketplace.

Context Engineering & Session Memory

Structured session memory and in-session reasoning that lets models keep track of ongoing tasks, enabling adaptive interactions that respond to user behavior over time.

Tool-Calling Layer

Complete tool-calling layer that captures model responses, invokes appropriate functions on the client device, and routes results back into the reasoning loop. Enables agentic workflows such as accessing real-time APIs, fetching local preferences, and invoking app-specific methods.

Resources