ChatGPT, Claude and other AI assistants can now work with Plota's planning data directly. Plota is an MCP server, the standard assistants use to call outside tools: add one address to your assistant, sign in with your Plota account and press Allow.
Several property data platforms have launched MCP servers this year. Most offer one as another way to call their API: the assistant makes calls, each call comes off a monthly allowance of tokens or credits, and the data stays on the platform. We didn't want to release an MCP until it felt truly native rather than a repackaged API. We took a lot of time to ensure it was fast and efficient.
A real answer, built from one call to Plota.
One call for everything about a site
A question about one property usually takes five or six lookups. investigate_site answers it in one call: applications at the site and beside it, the families they belong to, housing schemes of ten or more homes within a kilometre, the constraints at the point, and the council's decision times and approval rate.
The server also tells the assistant how planning records work. It follows the decision, since a council's status wording can lag behind it. Each answer cites the Plota page, which links to the council's own file, and where a council has decided too few applications for its figures to mean anything, the tool withholds them and says why.
Alerts from the conversation
Ask the assistant to keep an eye on a site, a street or a whole council and it saves a Plota alert. New applications then arrive by email, daily or weekly, or go straight to your own system by webhook. Later, "what has come in near my sites this week?" gets its answer from the alert's matches.
Anything that changes your account, such as saving an alert or starting an export, asks you to confirm first. A connection allowed read-only access cannot make changes.
Passing data to your other tools
Every reply carries the full records as structured data, with the same fields whichever council they come from, so an assistant with other connectors can pass them straight on. With BigQuery connected as well, "load every new-homes application in Cornwall this year into our planning table" is a single instruction. A Google Sheet or a CRM works the same way.
For bigger pulls, create_export returns a whole filtered set in one call, as JSON or CSV. On Pro, changed_since fetches only the records that changed since your last run, so a weekly refresh carries the changes and nothing else.
The rest of the API, as tools
Every endpoint in the Plota API is a tool as well, so a new endpoint becomes a tool the day it ships. Replies give the assistant one line per record and the full records once as data, which lets it read far more applications in one conversation. Applicant and agent names stay out of searches until you ask who applied, so browsing never spends your contact allowance. A council reference finds that application exactly, and where several councils use the same reference, the assistant is told to check which council you mean.
Every tool call is one API request on your plan, with no separate allowance for MCP, and investigate_site uses up to seven. Signing in replaces the API key, an email address with no account gets a free demo account as it connects, and you can see and end your connections from the API dashboard. Setup for ChatGPT, Claude, Claude Code and Cursor is on the MCP page.