For developers

Travel data your agent can actually call.

Sign up, copy a key, ship in 10 minutes. 44 grounded Concierge tools, first-class Python SDK, citations + license tag on every row. Pre-revenue and free tier; usage-based pricing only kicks in when you ship.

5-line itinerary planner.

No SDK install, no model selection, no chunking. One POST returns a grounded itinerary with per-POI citations.

# Try it right now — no API key required (10 free Concierge calls / hr / IP)
curl -sS https://api.travelminds.ai/v1/agent/concierge-public \
  -H "Content-Type: application/json" \
  -d '{"query": "4-day Buddhist pilgrimage from Tokyo, Japanese-speaking guide, mid-budget"}'

The response carries the itinerary plus a citations.data_sources block — every POI, every restaurant, every transit segment traces back to a verified source row. license_tag on each, so your compliance review takes minutes, not weeks.

Once you've got a key, the same call against /v1/agent/concierge unlocks the full 44-tool surface, higher per-minute limits, and per-tenant license filtering.

What changes: you stop hand-curating travel data per market. One API call replaces a weekend of scraping Wikipedia + Google Places + a Sabre integration that requires a sales call to even read the schema.
44
Grounded Concierge tools — graph traversals, POI lookups, calendar/currency utilities, license-posture filters. All composable in a single agent call.
Python SDK
First-class wheel + 3 working examples. pip install travelmindsai and you're calling tools in 90 seconds. JS / Go / Ruby on the roadmap.
99.5%
Monthly SLA on the cloud surface. 99.0% on-prem. Status page at /status with public incident history; no marketing-speak.

Why a tool registry instead of a flat REST schema?

Travel data is bigger than any single endpoint. You don't want to learn 60 routes; your agent should pick the right tool the way you'd pick the right library. The Concierge surface is router + 44 tools + structured grounding contract {answer, citations, license_tag, confidence} — every tool returns the same shape, so model context doesn't fragment.

Recall@8 on the embedding-based router is 0.98. Median LLM calls per query: 1. That means the agent spends tokens producing the answer, not searching for tools.

Free key, no credit card, 10 minutes to first call.

The Free tier ships 100 Concierge calls + 1,000 data calls per month. Upgrade when you ship.