Sign inTry it

Doctolib MCP server: let your AI agent search doctors and free slots

Wield is an MCP server that gives your AI agent Doctolib tools. Your agent can search practitioners by specialty and city, read free appointment slots for up to 7 days, and prepare a booking that waits for your click. The requests run inside your own browser with your own Doctolib session, so your login never leaves your computer.

Updated

Live recipedoctolib.de/get-slotsv1

Doctolib Get free appointment slots for a Doctolib practitioner

Read-only, no loginRuns in your browser only

Parameters (7)
agenda_idsinteger[]required
daysinteger
insurance_sectorstringMust match the patient's insurance; a wrong sector returns empty days, not an error "public" | "private"
practice_idintegerrequired
start_datestringrequiredYYYY-MM-DD
telehealthbooleanVideo consultation slots
visit_motive_idintegerrequired
Use it with your agent
Measured on the live site205× faster
MetricWield recipeBrowser Use + Claude Sonnet 5.5
Success100%50%
Median time345 ms71 s
Cost per run$0$0.164

8 runs each, 7 November 2026. How we measure

Wield recipes for Doctolibdoctolib.de
Everything for Doctolib

What your agent gets

Doctolib has no public API for patients. Wield's recipes use the same requests the Doctolib website makes when you search it yourself. In your agent, they show up as tools through wield-mcp:

RecipeWhat it doesLoginStatus (7 Nov 2026)
doctolib.de/search-practitionersDoctors and practices by specialty and city, for public (GKV) or private (PKV) insuranceNoVerified: 5 of 5 checks
doctolib.de/get-slotsFree slots for one practitioner and visit type, grouped by day, plus the next free slotNoVerified: 2 of 2 checks
doctolib.de/my-appointmentsYour upcoming appointmentsYesUntested, needs a logged-in recording
doctolib.de/book-slotSelects a slot in your account and returns the practice's noticesYesUntested write; always asks for your Allow

The search returns what the next step needs: name, specialty, address, whether new patients are accepted, video consultation, and the ids (visit_motive_id, agenda_ids, practice_id) that get-slots takes.

How a request flows

  1. You ask your agent: "Find a dermatologist in Berlin with a free slot this week, public insurance."
  2. The agent calls find_recipe, then run with doctolib.de/search-practitioners and {specialty: "Hautarzt", city: "Berlin"}.
  3. wield-mcp passes the run to the Wield extension through a local bridge. The extension sends the request from a doctolib.de tab, so it carries your normal browser session.
  4. The agent picks practitioners and calls doctolib.de/get-slots for each one.
  5. The results go back to your agent. Nothing passes through wield.run.

In our live check on 7 November 2026, a dermatologist search in Berlin returned 157 practitioners in 0.6 s over 3 requests. A slot lookup returned 32 slots over 7 days in 1 request.

Booking stays in your hands

book-slot is a write. Before any write request leaves your browser, Wield opens a window that shows the site, the task and every request with its body. Deny is the default. Two minutes without an answer also counts as a deny. The recipe never accepts a practice's notices for you and never confirms the appointment on its own: it hands you the next Doctolib page to finish yourself.

Doctolib's terms

Doctolib's terms prohibit automated access and scraping. The recipes are built for one lookup at a time that you start yourself, from your own browser. They are not for polling slots in a loop or collecting data in bulk. Health searches are sensitive, so results stay on your device and never appear in telemetry, which is off by default anyway.

FAQ

Is there an official Doctolib MCP server?

Not that we know of as of 7 November 2026.

Which AI apps does this work with?

Any MCP client. npx -y wield-mcp init sets up Claude Desktop, Claude Code, Cursor and Codex. Claude Desktop also has a one-click bundle on wield.run/install.

Does it work for Doctolib France or Italy?

Not yet. These recipes are for doctolib.de. You can request another site from your agent with request_recipe.

How fast is it compared with a browser agent?

On the slot task, the recipe passed 8 of 8 runs with a median of 0.35 s. Browser Use driving Claude Sonnet 5.5 passed 1 of 2 runs with a median of 71 s. Full method on the benchmark page.