Smyth Runtime Environment
The agent operating system for the agent economy
Stateful, secure, modular.
Meet SRE
Built from first principles, we designed an agent operating system to meet the needs of developers and enterprises.
Freely available to all humanity under the MIT license, and inspired by the elegance of the Linux architecture, we invite you to build the future with us.

Cross Platform
Runs anywhere: local, cloud, or edge. Fully cross platform: linux, mac, windows.
Observable
Built-in monitoring and debugging tools.
Modular
Swap any component without breaking the system.
Blazing Fast
Just 50MB, ultra-efficient architecture for multi agent orchestration.
Optimized for high-throughput AI operations.
Purpose Built
Designed from the ground up, specifically for AI agent workloads.
We built the kernel to power the internet of agents, and then open sourced it.
Security-First
Enterprise-grade security from day one. Done for you.


Set Up and Run AI Agents
Learn how to create and run AI agents using our CLI and SDK, and connect them to the MCP server for full integration.
Secure
Enough autonomy to matter, enough control to trust
OmniGuard
Intercepts each resource request and verifies embedded security tokens against agent permissions.
Key Vault
Your API keys are stored in an AES-256-encrypted vault. Secrets never touch disk or code, so credential leaks stop here.
Logs & Audit Trails
Detailed immutable logging of all agent actions and access attempts for compliance reporting, fine-tuning and incident investigation in Agent Studio.


Robust Debugging
Testing different use cases quickly and understanding exactly how the agent behaves is key to creating secure solutions.


Transparency
Security begins with understanding what your agent does. SmythOS can visualize more logic than any other platform.
Sandboxed
Components are fully sandboxed to prevent data leakage.
Hot Loading
Switch and add AI models on the fly without requiring a rebuild or redeploy.


Open Source
We have open sourced our runtime environment (SRE), command line (SRE CLI) and development kit (SRE SDK) under the MIT license.
We are preparing to open source our visual builder.
If you can’t contribute to the codebase, consider supporting our discord community.
Control
Safe agents are economically valuable agents
Aligned Automation
SmythOS architecture natively air gaps components and data flows on a “need to know” basis. Components are unaware of each other, act independently, with zero data leakage.


In this example, the AI model never gets to discriminate based on the person’s identity, and the patient privacy is respected.
Language Chains
Let one model refine the output of a previous AI model.


Agent Chains
Let one agent evaluate the output of another agent.


Human in the Loop
Let a human verify the output of an AI model or agent


Tree of Thought
Let one model refine outputs of several other AI models.


Reasoner Chains
Perform deep thinking from different models and perspectives, then evaluate.


Grounding
Augment AI reasoning with grounding over RAG, API, SQL and more.


Catch them All
Design agents the way you put human teams together for outcomes that are secure, grounded, reviewed, and refined.


Developers Love SmythOS
Join over 20,000 developers who build with Weaver.