Benchmark: get slots on Doctolib with an AI agent
On Doctolib, Wield's recipe passed 8 of 8 slot lookups with a median of 345 ms and no model cost per run. A browser agent running Claude Sonnet 5.5 passed 1 of 2 runs, with a median of 71 seconds and $0.16 per run. The agent side is small: one run found slots, and one returned no items.
Doctolib Get free appointment slots for a Doctolib practitioner
Read-only, no loginRuns in your browser only
Parameters (7)
| Metric | Wield recipe | Browser Use + Claude Sonnet 5.5 |
|---|---|---|
| Success | 100% | 50% |
| Median time | 345 ms | 71 s |
| Cost per run | $0 | $0.164 |
8 runs each, 7 November 2026. How we measure
The task
Both sides ran the same three slot lookups. Each query names one practitioner, a visit type, the agendas and a start date, all for public insurance.
| Query | Parameters |
|---|---|
| q1 | visit_motive_id = 16250139, agenda_ids = [2606844], practice_id = 815415, insurance_sector = public, start_date = 2026-11-09 |
| q2 | visit_motive_id = 1272266, agenda_ids = [208569, 208594, 208596, 425538, 875898, 937033, 1103620], practice_id = 82778, insurance_sector = public, start_date = 2026-11-10 |
| q3 | visit_motive_id = 15532055, agenda_ids = [2334737, 2478448], practice_id = 706045, insurance_sector = public, start_date = 2026-11-11 |
Pass check: at least one free slot, with a date and a time.
Results (7 Nov 2026)
| Wield recipe | Browser agent | |
|---|---|---|
| Runs passing the check | 8 of 8 | 1 of 2 |
| Median time per run | 345 ms | 70.6 s |
| Model cost per run | $0 | $0.16 |
| Requests per run (median) | 1 | 615 (28 to the site) |
| Blocked by the site | 0 runs | 0 of 2 runs |
The browser agent took a median of 11 steps, 44,972 input tokens and 2,319 output tokens per run. Its log records no HTTP 401, 403 or 429 response from the site. Wield had no failed runs. Its fastest run took 307 ms and its slowest 4,584 ms.
Why the results differ
The get-slots recipe takes the ids that a practitioner search returns, visit_motive_id, agenda_ids and practice_id. It then reads free slots for up to 7 days in one request. It runs from a doctolib.de tab in your browser, so it needs no login.
The browser agent has to open the practice page, pick a visit type and step through the calendar. Each of its median 11 steps waits for a model call. Free slots also change often, so each run shows what was free at the time it ran.
Caveats
- Small and preliminary. The benchmark was still running on 7 Nov 2026. The agent side has 2 runs. One returned no items, and the table counts it as not passing.
- Cost is model cost only. Wield's $0 covers model cost. No model runs in the loop when a recipe runs.
- Live numbers. The live benchmark row on this page comes from
bench/results.json.
Terms of Doctolib
Doctolib's terms prohibit automated access and scraping. The recipe runs one lookup that you start, from your own browser, with one request. Do not poll for slots. Polling is the pattern these terms are aimed at. Check the site's terms yourself. Wield's recipes note known restrictions.
FAQ
Does Doctolib block browser agents?
Not in these runs. The browser agent was blocked by the site in 0 of 2 runs, and its log records no 401, 403 or 429 response. One of its two runs returned no items, which is not a block. Other practitioners and dates may differ, and bot protection can change at any time.
How fast is the Wield recipe for get slots on Doctolib?
The median run took 345 ms over 8 runs. The fastest took 307 ms and the slowest 4,584 ms. All 8 runs passed the check.
Can I rerun this benchmark?
Yes. The method is on the benchmark page, and the query table above is the full input. Run it one lookup at a time. The raw per-run logs keep destinations, timing, tokens and checks. They do not keep page contents.