Tools Registry
AVAILABLEThe Tools Registry provides a unified catalog of executable functions, REST API endpoints, and webhook triggers exposed to your AI agents and models.
- Portal Location:
Projects > [Your Project] > AI & Capabilities > Tools(/projects/[slug]/tools) - API Endpoint:
/v1/tools
What is a Tool? #
A Tool follows the standard JSON Schema function-calling specification supported by models like GPT-4o, Claude 3.5 Sonnet, and DeepSeek:
- Name: Unique identifier (e.g.
get_weather,execute_sql). - Description: Clear explanation of when and why the model should invoke this tool.
- Parameters (JSON Schema): Argument names, types, constraints, and descriptions.
- Execution Handler: Webhook URL, local plugin worker, or external MCP server.
Rendering diagram...
graph LR
Model["Model generates Tool Call: execute_sql(query=...)"]
Model --> Gateway["Naagmani Tool Executor"]
Gateway --> CheckPolicy{"MCP / Tool Policy Allowed?"}
CheckPolicy -->|Yes| Execution["Run Target Webhook / Plugin"]
CheckPolicy -->|Require Approval| Admin["Human Approval Queue"]
Execution --> Return["Inject Tool Output into Context"]
Registering a Tool in Developer Portal #
- In your project workspace, click AI & Capabilities > Tools.
- Click Add Tool.
- Fill out:
- Tool Name: e.g.,
lookup_customer. - Description: "Retrieves customer profile and active subscriptions by email address."
- Parameters Schema (JSON):json
{ "type": "object", "properties": { "email": { "type": "string", "description": "The customer email address." } }, "required": ["email"] } - Target Endpoint: Webhook URL or MCP Server binding.
- Tool Name: e.g.,
- Click Save Tool.