AI Engineering
Build strict nested tool parameters once, add optional MCP output schemas and behavior hints, and export current OpenAI, Anthropic, or Model Context Protocol definitions.
Describe when the model should call the tool, its limits, and what it returns. Never put secrets or executable instructions in a schema.
Dot paths create nested objects; [] creates arrays.
| Path | Type | Required | Description | String enum | Actions |
|---|---|---|---|---|---|
Optional MCP structured-result schema.
| Path | Type | Required | Description | String enum | Actions |
|---|---|---|---|---|---|
MCP annotations are untrusted behavioral hints, not an authorization or confirmation mechanism.
{
"type": "function",
"name": "get_weather",
"description": "Get current weather for a city when the user asks about present conditions.",
"parameters": {
"type": "object",
"properties": {
"location": {
"type": "object",
"properties": {
"city": {
"type": "string",
"description": "City name"
},
"country_code": {
"type": "string",
"description": "Two-letter country code",
"enum": [
"US",
"CA",
"GB",
"IN"
]
}
},
"required": [
"city",
"country_code"
],
"additionalProperties": false
},
"units": {
"type": "string",
"description": "Temperature units",
"enum": [
"celsius",
"fahrenheit"
]
}
},
"required": [
"location"
],
"additionalProperties": false
},
"strict": true
}Formats follow official documentation checked on 2026-08-28. Revalidate provider constraints when integrating, and enforce permissions in the tool implementation.
Describe when the model should call the tool, its limits, and what it returns. Never put secrets or executable instructions in a schema.
Dot paths create nested objects; [] creates arrays.
| Path | Type | Required | Description | String enum | Actions |
|---|---|---|---|---|---|
Optional MCP structured-result schema.
| Path | Type | Required | Description | String enum | Actions |
|---|---|---|---|---|---|
MCP annotations are untrusted behavioral hints, not an authorization or confirmation mechanism.