Methodology 1.3.0 is live: transport from NaPTAN, amenities from a stored snapshot.Changelog
onegoodarea

For AI agents

Built to be read by AI.

OneGoodArea is infrastructure for AI property workflows, so we are built to be read, queried, and cited by machines, not only by people. Every entry point below is public, documented, and stable.

01 · Entry points

Everything an agent needs, in the open.

Four public, documented entry points. No login, no scraping, no guessing. Point a model at any of them and it can describe OneGoodArea correctly and link back to the source.

Also published: an llms-full.txt with the complete reference in one file, a discovery manifest at /.well-known/ai-plugin.json, a sitemap, a robots policy that explicitly welcomes AI crawlers, and schema.org structured data across the site.

llms.txt

/llms.txt. A curated map of the site for language models: what OneGoodArea is, the products, and where the canonical docs live.

MCP server

/docs/mcp. Query UK areas from inside AI tools and any MCP client. The agent calls the tool and gets structured data back.

OpenAPI spec

/openapi.json. The full REST API described in a machine-readable spec, so tools can generate clients and understand every endpoint.

Methodology

/methodology. How every score is built, weighted, versioned, and sourced. The reference an assistant needs to explain a number.

02 · Beyond reading

Not just readable. Callable.

OneGoodArea ships as a Model Context Protocol server, so an agent does not only read about the data, it runs the query and acts on the answer. A sample of the tools it gets:

score_postcode

Score any UK postcode for a chosen intent.

get_area_signals

Pull the underlying signals for an area, each with its source.

compare_postcodes

Compare two or more areas side by side.

find_areas

Find areas that match a set of criteria.

watch_portfolio

Watch a set of areas and track changes over time.

methodology_for

Return the exact methodology behind a score.

03 · Public and stable

Reproducible today, citable months from now.

Every response carries the engine version that produced it, currently v1.3.0. An answer an assistant gives today stays reproducible months from now, so it is safe to cite.

Building on UK area data?

If you are putting area intelligence inside a product or an agent, the founding pilot is the fastest way in.