Agentary

Agentary is an open-source, local-first framework designed for developers to create and deploy autonomous AI agents that run entirely within the user's browser. By utilizing technologies like WebGPU and WebAssembly, Agentary brings AI execution to the edge, eliminating the need for servers, cloud-based APIs, and external data processing. This approach prioritizes user privacy, reduces latency, and removes operational costs, allowing for the development of highly responsive and secure AI-powered web applications.

How to use Agentary

To get started with Agentary, developers can install the SDK via npm using the command npm install agentary-js. After installation, import the SDK into your JavaScript or TypeScript project. You can then define a goal for your AI agent, which can range from a simple prompt to a complex, multi-step workflow involving various tools. The framework handles loading a quantized language model into the browser, where it runs on WebGPU for accelerated performance. The agent then reasons, plans, and executes tasks locally, such as calling custom tools or APIs, without any data leaving the user's device.

Core Features of Agentary

• Browser-Native Execution: Runs quantized language models directly in the browser using WebGPU or WebAssembly, eliminating server dependencies.
• WebGPU Acceleration: Taps into the device's GPU for near-real-time inference and smooth, responsive token streaming.
• Agentic Workflows: Enables the creation and execution of complex, multi-step agent workflows with conditional logic.
• Tool & API Integration: Allows developers to register custom tools and API functions that the AI agents can call and utilize during execution.
• Multi-Model Support: Supports running different models in separate worker threads for parallel execution of tasks like chat, reasoning, and planning.
• Offline Functionality: Once models are downloaded, the agents can function completely offline, as all processing happens on the client-side.

Use Cases for Agentary

Agentary is ideal for applications where data privacy, low latency, and cost-efficiency are critical. Use cases include: Privacy-First Health & Wellness Companions that analyze wearable data locally; Creative Writing & Ideation Tools that operate without cloud API limits; Interactive Learning Tutors that keep student data private; Local Document Q&A for confidential research; On-Device Email & Writing Assistants for sensitive communications; and Embedded Customer Support Chatbots that reduce server costs while preserving user privacy.

Advantages of Agentary

The primary advantage of Agentary is its privacy-first, serverless architecture. By processing all data on the user's device, it guarantees that sensitive information is never exposed to third parties. This also results in significantly lower latency and a more responsive user experience. Furthermore, it eliminates server costs, API fees, and usage limits associated with cloud-based AI services. As a free and open-source tool under the MIT License, it offers developers complete control and flexibility for both personal and commercial projects.

Pricing and Plans

Agentary is completely free and open-source, distributed under the MIT License. There are no licensing fees, API costs, or usage limits, making it accessible for all personal and commercial projects.