Use with ChatGPT
Ask ChatGPT about
a Polish parcel.
Add landfeed as a connector for a remote MCP server. ChatGPT then answers parcel questions with one tool call, with a source and date on every fact.
Coming soon. The hosted server is not live yet. https://mcp.landfeed.pl/mcp is the planned address; join the waitlist to hear when it opens.
Add the connector
ChatGPT adds custom MCP servers as connectors. At the time of writing this needs developer mode, on the plans that offer it.
- Open Settings, then Apps & Connectors.
- Under Advanced settings, turn on Developer mode.
- Back in Apps & Connectors, choose Create.
- Name it
landfeed, and paste the MCP server URL:https://mcp.landfeed.pl/mcp - Choose OAuth for authentication, confirm that you trust the server, and choose Create. Sign in when asked.
Menu names are ChatGPT's, as of October 2026, and change often. In workspaces an admin may need to allow custom connectors first.
Use it in a chat
Start a chat, open the tools menu and pick landfeed (in developer mode the connector is listed there). Then ask your question. Every landfeed tool reads; none writes, so there is nothing for ChatGPT to confirm before it acts.
Example prompts
In English
- Is parcel 25/18 in Lesznowola buildable? Cite the sources.ChatGPT calls
check: verdict, named checks, sources and the record in one result. - List the parcels within 300 m of 141803_2.0006.25/18.ChatGPT calls
searchwith a radius: id, gmina, obręb, number, area and centroid for each. - What is the planning status of gmina Lesznowola?ChatGPT calls
layerwithplanningand only the gmina part is filled: plan ogólny status, MPZP count and coverage where known.
In Polish
-
PL pendingEnglish gloss: "Is plot 25/18 in Lesznowola buildable? Give the sources."
ChatGPT calls
check. -
PL pendingEnglish gloss: "Is this plot inside a protected area such as Natura 2000 or a landscape park?"
ChatGPT calls
layerwithprotected. -
PL pendingEnglish gloss: "Have any building permits been issued near this plot?"
ChatGPT calls
layerwithpermits: GUNB permits within 500 m, investor names never included.
What ChatGPT gets back
The same envelope as REST, as structured JSON. One layer answer looks like this:
{
"layer": "flood",
"coverage": "covered_no_data", // mapped area, parcel outside all zones
"value": { "q10": false, "q100": false, "q500": false },
"note": "…one sentence to quote…",
"source": { "title": "…", "attribution": "…", "licence": "…", "as_of": "…" },
"as_of": "…",
"layer_version": "…",
"hash": "…"
}
Limits to expect
- v1 holds mazowieckie only. Elsewhere the answer is
not_covered, refunded, andcorpus_coveragesays what is held. - Some layers are not loaded everywhere yet; they come back as
not_computedwith a null value, never a guess. - landfeed is not legal advice and never calls a parcel "safe".