Developers & agents
The Model Context Protocol server is how you read Propbar programmatically. This page covers how to connect to it, how authentication works, what happens when you hit a rate limit, and what an unauthenticated caller is entitled to see.
The recommended surface for AI assistants. Propbar's property research tools are exposed over the Model Context Protocol, so an assistant can search areas and properties, pull crime, schools, demographics and EPC data, and fetch comparables without you writing an integration.
https://mcp.propbar.co.uk/.well-known/oauth-protected-resource. Propbar renders its reports, maps and valuations from a property and area platform we license from Vepler. Bulk, commercial and high-volume access is granted directly on that platform under a separate written agreement; it is not part of the Propbar Service, and there is no self-serve REST tier.
Every error from api2.propbar.co.uk, including a wrong path, answers with a JSON body, never an HTML error page. Branch on code: it is stable, whereas error and message are prose and may be reworded.
{
"error": "Not Found",
"code": "NOT_FOUND",
"message": "Route /nonexistent not found",
"status": 404,
"docs": "https://propbar.co.uk/developers"
} Limits vary by endpoint and by the caller's tier, so read them from the response rather than hard-coding a number. A refused request answers 429 and carries:
Retry-After: seconds to wait before retrying. Honour it; retrying sooner will simply be refused again. RateLimit-Limit: the budget for the window that refused you.RateLimit-Remaining: 0 on a 429, by definition.RateLimit-Reset: seconds until the window resets.Retry-After is the standard field (RFC 9110) and the one to honour. The three RateLimit-* headers are the widely used names rather than a standard; they are sent for clients that already parse them.
Verified search-engine and AI crawlers are exempt from these limits when their identity confirms against the network their operator publishes.
Send Accept: text/markdown to this page, the homepage, /mcp or /property-data-api and you get the page as Markdown at the same URL, no HTML parsing required. Both variants send Vary: Accept.
/data-sources is HTML-only on purpose: its per-source licence and attribution text is the reason the page exists, and a summary must not stand in for it.
Crawlers and unauthenticated callers receive exactly what a signed-out visitor sees on the website. Paid depth requires an authenticated caller entitled to it: sale prices, comparables, the valuation and full risk detail. There is no crawler bypass.
Bulk and commercial access to the underlying platform is licensed separately; see Property Data API . Provenance and licensing for every field is listed on Data sources , and machine-readable crawling policy is at /robots.txt and /llms.txt.
It takes just a few seconds to get started.