AI Agents & Assistants
AVAILABLECreate, configure, and orchestrate autonomous AI agents equipped with multi-step reasoning, skill prompts, and tool execution bindings.
- Portal Location:
Projects > [Your Project] > AI & Capabilities > AI Agents(/projects/[slug]/agents) - API Endpoint:
/v1/agents
What is an AI Agent? #
An AI Agent in Naagmani combines:
- Core System Prompt: Base identity, behavioral guidelines, and system persona.
- Model Binding: Association with an underlying foundation model (e.g.
claude-3-5-sonnet) or dynamic Routing Policy (smart). - Skills: Attached instruction modules for specialized domain knowledge.
- Tools & MCP Servers: Registered functions allowing the agent to query databases, call APIs, or execute external tasks.
Rendering diagram...
graph TD
Agent["AI Agent (e.g. Support Assistant)"] --> Prompt["System Prompt & Personas"]
Agent --> Router["Model Router ('smart')"]
Agent --> Skills["Attached Skills (Refund Policy, Code Gen)"]
Agent --> Tools["Tools & MCP Connectors (SQL Query, GitHub)"]
Configuring an Agent in Developer Portal #
- Navigate to AI & Capabilities > AI Agents (
/projects/[slug]/agents). - Click Create Agent.
- Configure:
- Name & Description: e.g.,
Customer Triage Agent. - Base Model / Routing Alias: Select
gpt-4o,deepseek-chat, or virtual aliassmart. - Temperature & Max Tokens: Sampling hyperparameters.
- System Prompt: Core instruction template.
- Assigned Skills & Tools: Check which tools the agent is permitted to execute.
- Name & Description: e.g.,
- Click Save Agent. You can immediately test the agent in the Agent Playground.