Implementation:CrewAIInc CrewAI Tool Specs JSON

Knowledge Sources	CrewAI
Domains	Configuration, Tool Discovery, Schema
Last Updated	2026-02-11 00:00 GMT

Overview

tool.specs.json is a generated JSON specification file that comprehensively describes all available tools in the crewai-tools package, including their schemas, parameters, dependencies, and environment variable requirements.

Description

This file serves as a machine-readable catalog of the entire crewai-tools ecosystem. At over 26,000 lines, it covers every tool in the package with full schema information. The file is generated (not hand-written) by the generate_tool_specs.py utility.

The top-level structure is a JSON object with a single tools array. Each entry in the array is a tool specification object containing:

• name -- The class name (e.g., "AIMindTool", "SerperDevTool", "CodeInterpreterTool").
• humanized_name -- A display-friendly name (e.g., "AIMind Tool", "Serper Dev Tool").
• description -- Detailed description of the tool's purpose and capabilities.
• env_vars -- Array of environment variable definitions, each with name, description, required (boolean), and default (string or null).
• init_params_schema -- A JSON Schema / Pydantic schema for the constructor parameters, including type definitions, defaults, and descriptions.
• run_params_schema -- A JSON Schema for runtime execution parameters accepted by the _run method.
• package_dependencies -- Array of pip package names required by the tool (e.g., ["databricks-sdk"], ["stagehand<=0.5.9"]).

Usage

Use this file when building UI tooling, documentation generators, or platform features that need to dynamically discover and configure CrewAI tools without inspecting source code. It is also useful for validating tool configurations and generating type-safe API clients.

Code Reference

Source Location

• Repository: CrewAI
• File: lib/crewai-tools/tool.specs.json
• Lines: 1-26161

Structure

{
  "tools": [
    {
      "name": "AIMindTool",
      "humanized_name": "AIMind Tool",
      "description": "A wrapper around AI-Minds...",
      "env_vars": [
        {
          "name": "MINDS_API_KEY",
          "description": "API key for AI-Minds",
          "required": true,
          "default": null
        }
      ],
      "init_params_schema": {
        "properties": {
          "api_key": { "type": "string" },
          "datasources": { "type": "array" },
          "mind_name": { "type": "string" }
        }
      },
      "run_params_schema": {
        "properties": {
          "query": { "type": "string", "description": "..." }
        },
        "required": ["query"]
      },
      "package_dependencies": ["minds-sdk"]
    }
  ]
}

I/O Contract

Tool Specification Fields

Field	Type	Description
name	string	Python class name of the tool
humanized_name	string	Human-readable display name
description	string	Full description of the tool's purpose
env_vars	array	Environment variables required or optional for the tool
init_params_schema	object	JSON Schema for constructor parameters
run_params_schema	object	JSON Schema for runtime execution parameters
package_dependencies	array	List of pip package names required by the tool

Environment Variable Fields

Field	Type	Description
name	string	Environment variable name (e.g., "SERPER_API_KEY")
description	string	Human-readable description of the variable's purpose
required	boolean	Whether the variable is mandatory for tool operation
default	string or null	Default value if not set in the environment

Usage Examples

Reading Tool Specs Programmatically

import json

with open("lib/crewai-tools/tool.specs.json") as f:
    specs = json.load(f)

# List all available tools
for tool in specs["tools"]:
    print(f"{tool['humanized_name']}: {tool['description'][:80]}...")

# Find tools requiring a specific environment variable
serper_tools = [
    t for t in specs["tools"]
    if any(ev["name"] == "SERPER_API_KEY" for ev in t.get("env_vars", []))
]

# Get init parameters for a specific tool
tool_spec = next(t for t in specs["tools"] if t["name"] == "CodeInterpreterTool")
print(json.dumps(tool_spec["init_params_schema"], indent=2))

Related Pages

• Principle:CrewAIInc_CrewAI_Built_In_Tool_Selection