Recycling answers in your AI assistant
Last updated: 9 November 2026
Connect Dónde lo llevo to your AI assistant. It can then look up recycling guides and collection points in Spain, with their sources.
The connector is read-only. It needs no account, no sign-in and no API key.
How to connectOn this page
Connect your assistant
MCP (Model Context Protocol) is a standard way for an AI assistant to call tools on an outside service. This connector is an MCP server. Your assistant connects to it with the address below.
MCP server address
https://mcp.dondelollevo.com/mcpThis page sends no requests to the server. Copy the address or the command into your client.
Claude (web and desktop)
Go to Customize → Connectors → Add custom connector. Enter a name, for example
Dónde lo llevo, and the address above. There is no sign-in step. Then add the connector. See the Claude guide to custom connectors.Claude Code
Run this in your terminal:
claude mcp add --transport http dondelollevo https://mcp.dondelollevo.com/mcpCheck it with
claude mcp list. See MCP in Claude Code.Codex CLI
Run this in your terminal:
codex mcp add dondelollevo --url https://mcp.dondelollevo.com/mcpCheck it with
codex mcp list. See MCP in Codex.ChatGPT
Add the server from Plugins → + → Add custom MCP server, with a name, a description, the address above and authentication “none”. Then accept the notice, create the plugin and review the tools. See the ChatGPT connection guide.
Menus may differ by plan and client version. Your client may ask you to confirm before it uses an outside tool.
What you can ask
- “Where can I take used batteries in the Centro district of Madrid? Use Dónde lo llevo and cite the sources.”
- “What should I do with an old frying pan? Use Dónde lo llevo.”
- “Find clothing collection containers in Vigo with Dónde lo llevo, and tell me where the data comes from.”
The tool descriptions are in English. The data is in Spanish. The assistant answers in your language and keeps place names, limits and sources as given.
For a Madrid search, a district is enough. Do not give your home address or other personal data.
Coverage
- Guides: 60 recycling guides in Spanish. Each guide lists its sources and the municipal area each source applies to.
- Recycling points: fixed recycling centres, proximity points and mobile stops in the city of Madrid only, by district.
- Clothing collection points: Madrid, Barcelona, València, Zaragoza, Málaga, Vigo and Santa Cruz de Tenerife. Each city comes from its municipal open-data source, with its own licence and date. The tool returns the attribution, the licence and the caveats.
The connector does not compute the nearest point. It does not verify whether a point is open right now. Recycling rules and hours change: confirm with the municipal service before you travel.
Tool reference
There are six tools. All are read-only. Each tool rejects arguments that are not listed. A successful answer has result.isError: false. The data is in result.structuredContent, and the same data is repeated as JSON text in result.content[0].text. Most answers also have a message field with short instructions for the assistant.
The examples below are JSON-RPC 2.0 requests. JSON-RPC is the message format that MCP uses.
list_recycling_guides
Lists every published guide, grouped by theme, with its slug. A slug is the short identifier of a guide, for example pila. The assistant uses this list to map an object named in another language to its Spanish guide.
- Inputs
- None.
- Returns
total(number of guides) andgroups. Each group has atitleandguides, each withname,slug,titleandurl.
{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"list_recycling_guides","arguments":{}}}
search_recycling_guides
Finds the guide for an object or material by its Spanish name, for example sartén, pilas or ropa. It matches known object names only. It is not a full-text search.
- Inputs
query(required): string, 2 to 120 characters.- Returns
totalandguides: short summaries withname,slug,title,description,reviewed_onandurl.
{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"search_recycling_guides","arguments":{"query":"pilas"}}}
Shortened answer:
{"total":1,"guides":[{"name":"Pila doméstica","slug":"pila","reviewed_on":"2026-10-29","url":"https://dondelollevo.com/guias/pila"}],"message":"Call get_recycling_guide with this slug for the full guide."}
get_recycling_guide
Returns one full guide in Spanish. It includes the options for the object, steps, questions and answers, and the sources with the area each one covers. It also returns the acceptance rule of each kind of Madrid collection point.
- Inputs
slug(required): string, up to 100 characters, from the list or the search.- Returns
slug,name,title,url,reviewed_on,description,intro,collection_note,options(when,title,text),sections(heading,paragraphs,bullets,steps),questions(question,answer,source_url),sources(title,url,scope) andmadrid_acceptance_rules(point_kind,accepted,limit_text,source_url,checked_on).
A municipal rule applies to the area of its source. It does not apply to all of Spain. reviewed_on is the date of the last review of the guide.
{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"get_recycling_guide","arguments":{"slug":"menaje-metalico"}}}
find_recycling_points
Lists the published collection points of one district of the city of Madrid: fixed recycling centres, proximity points and mobile stops, with address, hours and source. With object_slug, it returns only points without a closure notice whose kind accepts that object, with the quantity limit.
- Inputs
district(required): one of the 21 Madrid district identifiers below.object_slug(optional): a guide slug, to keep only the points that accept that object.kind(optional):fixed,proximityormobile.offset(optional): thenext_offsetof the previous answer, to read the next page.- Returns
coverage,district,district_url,guide_url,total,offset,next_offset(empty on the last page),points(up to 20 per page) andacceptance_rules. Each point hasid,kind,kind_label,name,address,district,schedule_text,weekly_hours,closure_notice,acceptance_limit,acceptance_source_url,source_url,source_dateandurl.
The 21 values of district
arganzuela, barajas, carabanchel, centro, chamartin, chamberi, ciudad-lineal, fuencarral-el-pardo, hortaleza, latina, moncloa-aravaca, moratalaz, puente-de-vallecas, retiro, salamanca, san-blas-canillejas, tetuan, usera, vicalvaro, villa-de-vallecas, villaverde
{"jsonrpc":"2.0","id":4,"method":"tools/call","params":{"name":"find_recycling_points","arguments":{"district":"centro","object_slug":"pila"}}}
Acceptance is documented per kind of point. It does not come from an inspection of each point. An empty list does not mean there is no alternative.
get_recycling_point
Returns one Madrid collection point by its id. It includes the address, coordinates, weekly hours, closure notice, phone, source, and the objects its kind accepts with their limits.
- Inputs
id(required): integer, the point id fromfind_recycling_points.- Returns
id,kind,kind_label,name,address,district,neighbourhood,latitude,longitude,phone,schedule_text,weekly_hours(weekday1 = Monday to 7 = Sunday,opens,closes),closure_notice,accepted_objects(object_slug,name,limit_text,guide_url,source_url),source_url,source_date,checked_onandurl.
{"jsonrpc":"2.0","id":5,"method":"tools/call","params":{"name":"get_recycling_point","arguments":{"id":1843}}}
Hours come from the municipal source. They are not verified in real time.
find_clothing_points
Lists used-clothing containers and collection points in one city of the clothing map. Each city comes from its municipal open-data source, with its own date, licence and caveats.
- Inputs
city(required):madrid,barcelona,valencia,zaragoza,malaga,vigoorsanta-cruz-de-tenerife.area(optional): 2 to 80 characters of text to match in the district, neighbourhood or street, as written in Spanish.offset(optional): thenext_offsetof the previous answer.- Returns
city(name,attribution,licence_label,licence_url,source_url,obtained,source_note,finder_url),map_url,total,offset,next_offsetandpoints(up to 20 per page). Each point hasid,kind,name,address,area,latitude,longitude,acceptance,hours,operator,recycling_point_id,source_urlandsource_updated.
{"jsonrpc":"2.0","id":6,"method":"tools/call","params":{"name":"find_clothing_points","arguments":{"city":"vigo"}}}
Shortened answer:
{"city":{"name":"Vigo","attribution":"Concello de Vigo","licence_label":"ODC-BY 1.0","source_url":"https://datos-ckan.vigo.org/dataset/lp-reciclaje-textil"},"total":308,"offset":0,"next_offset":20,"points":[…]}
The assistant should name the attribution and the caveats in source_note.
Limits and errors
- Usage limit
- 500 tool calls per 24 hours per source address. The window starts at the first call.
- Assistant platforms
- Many people reach the connector from the same addresses of an assistant platform. Calls from the address ranges of assistant platforms share 20,000 calls per 24 hours per range.
- Limit reached
- The tool answers with an error message that says how long to wait. The HTTP status stays 200.
- Counters
- The counters live in memory only.
- Tool errors
- Wrong arguments, an unknown guide, point, district or city, and unavailable data come back as a tool error:
result.isError: truewith a short English message inresult.content[0].text. Example:Unknown guide. Use a slug from list_recycling_guides. - Protocol errors
- An unknown tool or method comes back as a JSON-RPC
errorobject. Examples:-32602 Unknown tooland-32601 Method not found. - No results
- A correct answer can have
total: 0. It does not prove that no option exists. Read the guide or ask the municipal service.
Technical details
- Endpoint:
https://mcp.dondelollevo.com/mcp - Transport: Streamable HTTP. The client sends
POSTrequests and receives JSON answers. - GET: returns HTTP 405 by design. If you open the address in a browser, you see this error. It is expected.
- Authentication: none. No account, no OAuth, no API key.
- Protocol versions:
2025-11-25and2025-06-18. The version is negotiated ininitialize. - Request body: 16 KB at most.
- Access: read-only. No tool changes data.
Privacy and help
The service receives the tool arguments and the IP address of the caller. That address is normally the address of the AI platform, not of your device. It does not receive your conversation. Do not put personal data in your questions.
Read the privacy policy and the terms of use. For questions or to report an error, use the contact page or write to contactlollevo@gmail.com.