Tools allow AI models and autonomous agents to interact with external databases, APIs, and microservices via standard JSON Schema function calling.

Portal Page: /projects/[projectId]/tools

Rendering diagram...

graph TD
    LLM["Model (Needs Customer Data)"] -->|tool_call: get_customer_by_id| GW["Naagmani Tool Engine"]
    GW --> Validate{"Validate Parameters against JSON Schema"}
    Validate -->|Valid| Execute["Execute Target HTTP / gRPC Endpoint"]
    Validate -->|Invalid| Error["Return Schema Validation Error to Model"]
    Execute --> Result["Inject Tool Output into Conversation Context"]

Tool Definition Format #

Every tool in Naagmani is defined with standard JSON Schema parameter specifications:

json

{
  "name": "fetch_user_account",
  "description": "Fetches a user account and subscription status by email.",
  "parameters": {
    "type": "object",
    "properties": {
      "email": {
        "type": "string",
        "format": "email",
        "description": "The user's registered email address."
      },
      "include_billing": {
        "type": "boolean",
        "description": "Whether to return active subscription and invoice history."
      }
    },
    "required": ["email"]
  },
  "endpoint": {
    "url": "https://internal-api.example.com/v1/users",
    "method": "POST"
  }
}

Next Steps #

Integrate Model Context Protocol: MCP Servers & Transports
Enforce tool access policies: MCP Tool Policies

Tools Architecture & Working Flow

Tool Definition Format #

Next Steps #