SWAI Kinesis
The action layer for AI agents
Kinesis is SWAI’s execution layer — enabling agents to interact with external systems and trigger real-world outcomes.
Like acquiring new skills, Kinesis allows agents to expand their capabilities without rewriting core logic.
These skills allow agents to connect with external platforms, interact with web services, and communicate with users.
Connection to Cortex
Kinesis works in tandem with Cortex, the decision-making brain of SWAI agents. When Cortex determines an action should be taken, Kinesis executes that decision through its skill system.
Kinesis and Cortex operate in a loop: Cortex makes decisions, Kinesis executes them — and then sends performance feedback back for ongoing adaptation.
Capability Awareness
Kinesis informs Cortex about available skills, ensuring the agent only attempts actions it can actually perform
Action Execution
When Cortex decides on a course of action, Kinesis translates that decision into concrete steps
Feedback Loop
After executing actions, Kinesis reports results back to Cortex, allowing for learning and adaptation
This skill-based architecture enables agents to develop specialized capabilities while maintaining a coherent decision-making process.
As new skills are added to an agent’s Kinesis module, its action possibilities expand automatically without requiring changes to the core reasoning system.
AI Models Integration
Can integrate with domain-specific models (e.g. audio, vision, LLMs) to expand action types and precision
Agentic Frameworks
Connecting with Eliza and other agent ecosystems
Direct Actions
Executing external operations like API calls, web interactions, and on-chain actions
Internal Actions
Managing memory operations, state transitions, and self-monitoring processes
Creating Agent Swarms
One of the most powerful aspects of Kinesis is its ability to integrate with other specialized agents and AI models.
This capability enables the creation of agent swarms - collaborative networks where multiple agents with different skill sets and configurations work together seamlessly.
In these swarms, each agent contributes its unique capabilities and expertise, handling specific tasks according to its configuration.
This distributed approach to problem-solving creates a collective intelligence that is significantly more effective and versatile than any individual agent could be on its own.
Agent swarms enable problem-solving at scale — where each agent contributes specialised actions to move the system forward