SmythOS provides an operating-system-style infrastructure for agentic AI, with the core being the Smyth Runtime Environment (SRE) kernel, SDK, and CLI. Its goal is to make the building and deployment of production-grade AI agents simpler, while balancing autonomy with controllability, built-in security, and open accessibility.
Purpose and Positioning SmythOS is described as "Linux for AI agents," inspired by operating system kernel architecture, providing a scalable foundation for agent orchestration and lifecycle management. Developers can use it to build, run, and manage agents, and deploy them across local, cloud, and edge environments. If a visual drag-and-drop interface is preferred, its open-source visual agent studio, SmythOS Visual Agent Studio, can be used.
Main Capabilities
- Unified resource abstraction: Provides a unified interface for resources such as LLMs, vector databases, storage, and caching, with the same set of APIs remaining consistent across all providers. Switching providers (for example, storage from local to S3, or LLM from OpenAI to Anthropic) requires no changes to business logic.
- Agent-First design: Built specifically for AI agent workloads.
- Modular architecture: A pluggable connector system that supports extension to any infrastructure.
- Production-ready: Scalable, observable, and battle-tested.
- Enterprise-grade security: Built-in access control and secure credential management, requiring every operation to receive the corresponding authorization through the Candidate/ACL system.
- 40+ production-ready components: Covering scenarios such as AI/LLM, external calls, data processing, logic, storage, and code, callable programmatically or through symbolic workflows in .smyth files.
- Supported connectors: Storage (local, S3, Google Cloud, Azure), LLM (OpenAI, Anthropic, Google AI, AWS Bedrock, Groq, Perplexity), vector databases (Pinecone, Milvus, RAMVec), caching (RAM, Redis), Vault (JSON files, AWS Secrets Manager, HashiCorp).
Usage The repository is a monorepo containing three main packages: the SRE core runtime (packages/core), the SDK (packages/sdk), and the CLI (packages/cli). Two installation methods are provided: Method 1 (recommended): Install the CLI globally and create a new project. The CLI will guide users step by step in creating an SDK project configuration suited to their needs. Method 2: Add the SDK directly to an existing project. The SDK allows agents to be built in code, or .smyth files to be loaded and run (.smyth is the agent extension generated by the SmythOS builder). The SDK supports stream mode and chat mode. Examples are provided in the documentation, such as loading and running a .smyth file, and using SDK elements to write the logic for an "article writing agent."
Component System 40+ components cover a variety of uses, with examples including:
- AI/LLM: GenAILLM, ImageGen, LLMAssistant
- External: APICall, WebSearch, WebScrape, HuggingFace
- Data: DataSourceIndexer, DataSourceLookup, JSONFilter
- Logic: LogicAND, LogicOR, Classifier, ForEach
- Storage: LocalStorage, S3
- Code: ECMAScript, ServerlessCode
Applicable Scenarios Suitable for developers and teams that need to move AI agents from development to production, especially in scenarios where they want flexibility to switch between different providers such as LLMs, storage, and vector databases, need enterprise-grade security and data isolation, or want to run agents uniformly across local, cloud, and edge environments.
Limitations and Notes The project uses the MIT license. The official statement says that an open-source visual agent IDE will be released later this year. Information such as pricing, fees, platform support details, and performance data is not yet provided in the materials. If problems occur with the CLI or code, you can set the environment variable LOG_LEVEL="debug" and run it again, then share the logs for diagnosis.
