Core Concepts
Coralized Agents
Coralized Agents are autonomous, modular services that follow a shared interface, allowing them to participate in collaborative workflows within the Coral Protocol. They can reason, take action, and interact with other agents in a structured and secure way.
Unlike monolithic models or single-purpose scripts, Coralized Agents are designed to:
- Operate independently while coordinating with other agents when needed
- Represent well-scoped capabilities that can be composed into broader systems
- Communicate and reason within structured task threads
They follow a standard interface and behaviour contract as defined by the Model Context Protocol (MCP), enabling seamless interoperability across the ecosystem.
Core Characteristics
Each Coralized Agent is:
Lifecycle of a Coralized Agent
- Coralization Developers register external models, scripts, or services into the Coral Ecosystem using Coralizer modules. This process maps external capabilities to the Coral format.
- Registration The agent advertises its name, role, capabilities, and memory settings to the Coral Server. It becomes discoverable and callable within the ecosystem.
- Participation When invoked or mentioned in a thread, the agent processes incoming messages and executes its defined logic. It can:
- Respond directly
- Perform tool invocations
- Mention or delegate to other agents
- Deactivation After a task is complete or the thread is closed, the agent’s temporary state is cleared (unless long-term memory is explicitly enabled).