Agents Overview
Agents act as orchestrators within the C3 Agentic AI Platform and integrate with your existing systems. For example, you can configure agents to connect to CRMs or data lakes like Snowflake for automated information retrieval, execute multi-step workflows by calling specialized tools in sequence, and maintain conversation context to provide coherent responses across extended interactions. When deployed to production, agents scale automatically to handle concurrent requests.
Agent capabilities
Agents are intelligent entities that can:
- Understand and respond to natural language questions and requests.
- Access external systems through custom tools and integrations.
- Remember context across conversations within a session.
- Execute workflows by calling multiple tools in sequence.
- Stream responses for real-time user interfaces.
- Scale automatically when deployed to production environments.
Core agent Types
The platform provides several core Types for building and managing agents:
- GenaiCore.Agent: Represents a single agent configuration with defined behavior and capabilities
- GenaiCore.Agent.Deployment: A deployed agent running persistently in production, accessible to end users
- GenaiCore.Agent.Draft: Editable agent configurations you can modify and test before deployment
- GenaiCore.Agent.Template: Pre-built agent configurations that can be duplicated and customized
- GenaiCore.Agent.Run: Records of individual conversations, including inputs, outputs, and execution status
- GenaiCore.Agent.Message: Enables agents to emit progress or output messages, accessible through streaming APIs
- GenaiCore.Tool: Custom functions that extend agent capabilities beyond text generation
- GenaiCore.SessionStore: Manages conversation memory and state across user interactions
Create agents
You can create agents in two ways:
- From scratch: Build a completely custom agent using the visual workbench. See Create Agents from Scratch.
- Duplicate agents: Modify pre-built agent configurations from Templates for common scenarios, or create copies of agents from Store or Drafts to create variations or customize agents created by other users. See Duplicate Agents.
Configure agents
Agents require configuration in four areas: objectives and prompts, model selection, memory settings, and tools. The Agent Workbench provides a visual interface for all configuration tasks. See Configure Agents for detailed configuration instructions.
For information about extending agent capabilities with tools, see Tools Overview.
Test and debug agents
Test agent responses in real-time using the chat panel in the Agent Workbench. Review detailed execution information through traces that show operation sequences, timing, and data flow. See Debug using Traces and Spans for debugging guidance and Observability and Monitoring Overview for monitoring deployed agents.
Deploy agents
After configuration and testing, move agents to Store for organizational sharing or deploy them for end-user access. Configure deployment settings including node pool, thread pool size, and tracing mode. Monitor deployed agents through metrics, traces, and deployment history. See Deploy Agents for deployment instructions and Agent Deployment Management for managing deployed agents.