getstrait 0.1.0

Python SDK for the Strait hotel booking API — booking infrastructure for AI agents.

Strait Python SDK

Hotel booking for AI agents. Python client for the Strait API.

pip install getstrait

• Resolve a hotel name to a property ID
• Check availability and pricing
• Book with human-in-the-loop approval (or skip it for autonomous flows)
• Cancel confirmed bookings
• One-line tool adapters for Anthropic and OpenAI

Full API reference: getstrait.dev

---

Auth

Pass api_key= or set STRAIT_API_KEY:

from strait import Strait

strait = Strait(api_key="sk_live_...")       # explicit
strait = Strait()                            # picks up STRAIT_API_KEY

Override the base URL (staging, self-host) with base_url= or STRAIT_BASE_URL.

---

Agent loop — Anthropic

Drop Strait into an Anthropic agent in three lines: get the tool definitions, run the model, and dispatch tool calls back through execute().

import anthropic
from strait import Strait

strait = Strait()
client = anthropic.Anthropic()

messages = [{"role": "user", "content": "Book me a night at Hotel Berlin on April 10."}]

while True:
    resp = client.messages.create(
        model="claude-opus-4-6",
        max_tokens=1024,
        tools=strait.tools.for_anthropic(),
        messages=messages,
    )
    if resp.stop_reason != "tool_use":
        break

    tool_results = []
    for block in resp.content:
        if block.type == "tool_use":
            result = strait.tools.execute(block.name, block.input)
            tool_results.append({
                "type": "tool_result",
                "tool_use_id": block.id,
                "content": str(result),
            })
    messages += [
        {"role": "assistant", "content": resp.content},
        {"role": "user", "content": tool_results},
    ]

By default, hotels_book and hotels_cancel_booking return UI-render data so a human can approve. Pass mode="auto" in the tool input to skip approval.

---

Agent loop — OpenAI

from openai import OpenAI
from strait import Strait

strait = Strait()
openai = OpenAI()

resp = openai.chat.completions.create(
    model="gpt-4.1",
    tools=strait.tools.for_openai(),
    messages=[{"role": "user", "content": "Find me a hotel in Berlin for April 10-12."}],
)

Dispatch tool calls the same way: strait.tools.execute(name, json.loads(args)).

---

Direct API

The SDK also exposes the raw resources if you'd rather call endpoints yourself:

from strait import Strait

strait = Strait()

# 1. Resolve
match = strait.properties.resolve(name="Hotel Berlin", city="Berlin")
property_id = match.results[0].property_id

# 2. Availability
avail = strait.properties.availability(
    property_id=property_id,
    check_in="2026-04-10",
    check_out="2026-04-12",
    guests=2,
)
offer = avail.available_offers[0]

# 3. Book — auto mode (no human approval UI)
booking = strait.checkout.book(
    offer_id=offer.offer_id,
    guest_name="Jane Doe",
    guest_email="[e-mail]",
    mode="auto",
)
print(booking.confirmation_code)

Human approval vs auto

checkout.book() returns either a SessionResponse or ExecuteResponse depending on mode:

mode	returns	side effects
"human_approval" (default)	SessionResponse	Creates a checkout session; UI renders ui_render_data.
"auto"	ExecuteResponse	Creates session and executes the booking immediately.

Use the session pattern when a human needs to see the price before charging — your frontend renders session.ui_render_data and confirms via your own UI flow. Use "auto" for unattended agents.

Cancellation

# Preview (returns UI render data; does not cancel)
details = strait.bookings.cancel(confirmation_code="RKCRQJDV")

# Confirm cancellation
result = strait.bookings.cancel(
    confirmation_code="RKCRQJDV",
    mode="auto",
    reason="plans changed",
)

---

Errors

All errors subclass StraitError:

from strait import (
    APIError,              # non-2xx response (base)
    AuthenticationError,   # 401
    NotFoundError,         # 404
    RateLimitError,        # 429
    ServerError,           # 5xx
)

try:
    strait.properties.availability(property_id="bad", check_in="...", check_out="...", guests=2)
except NotFoundError as e:
    print(e.status_code, e.body)

---

Notes

• Sync only for now. An AsyncStrait client is on the roadmap — raise an issue if you need it sooner.
• Python 3.10+.
• Use as a context manager to close the HTTP pool deterministically:

  with Strait() as strait:
      ...
  

---

License

MIT.