Agentic Lifecycle Management
In the C3 Agentic AI Platform, Agentic Lifecycle Management refers to the integrated system for developing, testing, deploying, and monitoring AI agents in a C3 AI application.
An agent is an autonomous entity within an application that can:
- Understand and respond to natural language queries.
- Use tools in workflows to accomplish designated tasks.
Agentic lifecycle management supports both UI-driven and developer-driven workflows.
Your application requires a genaiPlatform package dependency to create, modify, or deploy agents in C3 AI Studio. The genaiPlatform package provides the building blocks, APIs, and user interfaces for creating AI-powered applications and agents.
Components of agentic lifecycle management
The key components of Agentic Lifecycle Management are:
The components are shown in the following diagram:
The platform uses an agent invocation framework that abstracts agent implementation details. Each agent accepts context and input, and returns a stream of events. The architecture is framework-agnostic and supports both stateless and stateful (conversational) agents.
Development and testing
You can create and configure agents in C3 AI Studio using a visual Agent Workbench, or through code using frameworks such as LangChain or Autogen. Select prompts, context, and tools; and test agents interactively through a chat interface. The interface provides traceability for agent execution.
During development, you can iterate on an agent’s prompt, context, or tool selection. Test agent responses and refine configurations using the embedded chat interface in the Agent Workbench. Save an agent as a Draft to continue refining its behavior and outputs.
After development, you can publish agents and tools to a registry for reuse and sharing. An agent in an application Store is ready for deployment and cannot be directly modified. However, users with application access can duplicate stored agents for further customization.
To learn about the agent creation user interface, see Agent Gallery in C3 AI Studio. For a walkthrough on how to create an agent in C3 AI Studio, see Create Agents.
Deployment
Application administrators can deploy, enable, disable, and configure agents without code changes or downtime. Agent deployment is decoupled from application code, allowing dynamic agent selection and configuration at runtime. C3 AI Studio provides the following deployment options:
- Deploy the agent to your application: Make an agent available to other users with access to your application.
- Embed the agent in the application package: Make an agent available wherever your application package is used as a dependency.
Deployed agents appear in the Agents > Deployments section of C3 AI Studio.
Observability and monitoring
After deployment, you can monitor an agent’s performance to guide future iterations or development efforts. Observability provides insight into agent performance and response times during message processing.
The platform provides metrics such as token usage, response duration, and workflow bottlenecks. This enables proactive issue detection and continuous improvement based on real-world usage and feedback.
For guidance on monitoring deployed agents, see Observability and Monitoring Overview.