API reference · Natural Language Scheduling
Natural Language Scheduling
Parse natural language into a structured event, with per-field confidence.
★
Aspirational surface. This resource is part of the developer-platform roadmap. The field shapes follow the same conventions as live resources and are stable for prototype integrations.
Endpoints
POST
/v1/scheduling/parse
Natural-language → event
Parse natural language into an event
HTTP
POST /v1/scheduling/parse
Returns a fully-typed Event candidate plus per-field confidence scores. Drop the event object straight into POST /v1/events.
JSON
{
"text": "Lunch with Sarah Tuesday at 12 at Buena Vista Cafe",
"user_context": {
"timezone": "America/Los_Angeles",
"user_id": "usr_1abc2def3ghi",
"default_event_duration_min": 60
}
}Response:
JSON
{
"object": "parsed_event",
"event": {
"title": "Lunch with Sarah",
"start_time": "2026-05-26T12:00:00-07:00",
"end_time": "2026-05-26T13:00:00-07:00",
"timezone": "America/Los_Angeles",
"location": {
"name": "Buena Vista Cafe",
"address": "2765 Hyde St, San Francisco, CA 94109"
},
"attendees": [
{ "name": "Sarah", "email": null }
]
},
"confidence": {
"title": 0.96,
"start_time": 0.91,
"location": 0.88,
"attendees": 0.62
},
"warnings": ["attendee_email_missing"]
}Handling low confidence
We recommend surfacing a confirmation step to the user whenever any top-level field's confidence is below 0.7. The parser is conservative — it would rather fail-soft than fabricate.