A New Front Door for Hotel Bookings
Fliggy, the travel division of Alibaba Group, has opened its travel inventory to AI agents through an open-source skill called flyai. The repository sits at github.com/alibaba-flyai/flyai-skill under an MIT licence, dates its first commit to 19 March 2026, and held more than 1,100 stars at this reading.
Fliggy confirmed on 25 March 2026 that its AI interface now processes bookings. Planning support came first, with the AskMe assistant and the AliBtrip corporate tool in 2025.
The booking layer runs on Alibaba's Qwen app and covers the hotel transaction.
What the Skill Returns
The skill documents eight commands: a keyword search, a semantic search, and structured searches for flights, trains, hotels, attractions and Marriott Group hotels and packages. A hotel search takes a destination name and returns bookable results with direct booking links.
The documented parameter set for a hotel search is a destination, optional key words, a nearby point of interest, a hotel type of hotel, homestay or inn, a sort order, a check-in date, a check-out date, a star rating, a bed type and a nightly price ceiling.
None of the documented flags covers a part-day window. The hotel search carries no arrival-hour flag, no departure-hour flag and no duration in hours.
A day-use seller therefore needs a route into the agent that the overnight search does not supply.
The Transaction Layer Already Works
Fliggy's March release reports the volumes behind the interface. During the Chinese Spring Festival campaign, Fliggy's overall AI orders rose 800 per cent year on year. Attraction ticket orders rose more than 24-fold and were the most popular category. Conversion rates on air, train and attraction tickets reached the levels of established online travel agencies.
The Qwen app recorded more than 200 million AI-native shopping orders in the same campaign.
Dr Alex Chen, chief technology officer at Fliggy, described the shift in his own words. After the planning products of 2025, he said, the AI interface has moved closer to fulfilment. He said Qwen now connects Fliggy, Taobao, Taobao Instant Commerce, Alipay and Amap, and that consumers ask in natural language for a booking to be completed on their behalf.
The 800 per cent figure covers Fliggy's AI orders across all categories.
Where the Skill Ships
flyai is published on ClawHub and on GitHub. ChinaTravelNews reported more than 100 stars within 24 hours of the launch.
Fliggy's launch coverage lists the smartphone skill stores of Huawei, Xiaomi, OPPO and Honor among the device ecosystems the product targets.
The skill runs inside Claude Code, OpenClaw and other skill-compatible agents, needs no account setup and no API key to start, and returns results as single-line JSON.
What a Day-Use Seller Does With This
An agent can sell the inventory it can read. The documented hotel search reads a destination and a date range, so a daytime window has no place in the request.
Two checks follow for an operator holding day-use inventory. First, whether the day-use product sits in an agent-readable catalogue with a stated start hour, end hour and price. Second, whether the property's own booking page states those same three fields for a text-based assistant to read.
The Qwen order volumes show the route works for overnight rooms and attraction tickets. The day-use product needs the same structured treatment before an agent can sell it.