# Climate Insights Directory

> A public index of climate research openly published by civil society organisations, built to be searched and cited by both people and AI agents. The directory improves the discoverability of research that is already openly available; it does not publish it. Every record points back to the organisation that originally published the document.

The mission is to improve the discoverability of openly available, evidence-based insights on the climate crisis for the general public, academics, researchers, journalists, policy makers and others. The corpus is an index of reports, policy briefs, analyses and other documents openly published by civil society organisations worldwide. Government documents and academic literature have large-scale knowledge management systems; civil society's backfile of reports and briefings does not, and this directory exists to close that gap.

The directory is hosted by the Meliore Foundation and built by a network of climate communications researchers. The corpus, the document collection system and the classification methodology are developed with DiscourseLab, the project's data partner, whose social science research on climate communication underpins the taxonomies the platform runs on. Alongside search, content pages carry Schema.org JSON-LD and are indexed by a sitemap; a public JSON search API is available at `/api/v1/search`.

## Scope

The directory indexes openly published content from civil society. It does not intend to index governmental, academic or news media and journalistic content at this time. It is an index rather than a curated collection built for a particular audience or sector, so it aims for breadth: organisations of all sizes and mission scopes, older publications as well as recent ones, and evidence-based policy statements as well as full reports — while excluding event pages and press releases.

As a matter of climate information integrity, the directory excludes, as far as possible, content from organisations funded by the extractive or polluting industries the IPCC describes as major contributors to climate change, and organisations identified by researchers in the CAAD network as associated with climate denial or misinformation.

The corpus grows through releases of 50–100 organisations at a time, chosen to fill topical and geographical gaps. Absence of an organisation indicates that the editorial process has not reached it yet; it is not a judgement about that organisation. Each release is described on the Updates log.

## Metadata and reliability

Some fields are extracted directly from the original documents and are of high quality, particularly where the source formatting is consistent: author, organisation name, title and publication date.

Other fields are produced by an experimental labelling process using small, locally run language models, checked against climate-specific vocabularies and quality control measures. These are published in "Alpha, still testing" status: document summaries, highlighted insights, regions discussed, and classifications of climate topic, framing and argumentation type. Topic tags are still being normalised against a controlled vocabulary, and a methodology will be published once the process stabilises. Treat these fields as provisional and do not rely on them as authoritative descriptions of a document; the extracted fields and the document itself are the firmer ground.

## Access, citation and record stability

Document records contain metadata and summaries about the original document. The directory is not a substitute for the original document , and each record includes a direct link to where the publication was found. When citing research, prefer the original publisher's URL where the record carries one; directory record URLs should also be included as a "found via" part of a citation.

Individual records live under `/document/{id}`. The ids have an organizational identifier (up to 8 letters lower case) and a document number (up to 6 numeric digits). The document id should be stable, and should be referred to in citations with a CID prefix or a link to the document.

A record that has been withdrawn or editorially hidden answers 410 Gone, not 404, so a previously cited record can be told apart from a bad URL. Any organisation may request removal of its content from the directory by writing to info@climateinsightsdirectory.org.

Nine routes negotiate Markdown content when requested with an `Accept: text/markdown` header: `/`, `/about-us`, `/terms-of-use`, `/privacy-policy`, `/cookie-policy`, `/ai-transparency`, `/document/{id}`, `/geography/{slug}`, and `/geography/{geography}/{topic}`.

## Concept pages

Concept pages at `/geography/{slug}` and `/geography/{geography}/{topic}` gather civil-society research on a region or topic. Each concept page holds an editorial Overview and questions answered by insights that cite documents. Every published concept page is listed in `/sitemap-institutional.xml`.

## Search syntax

Search accepts field operators typed inside `q`, and filters passed as separate URL parameters. Everything combines with AND: each operator, filter and free-text term narrows the result set further.

Eight field operators exist. They are written as `name:value` inside `q` and are removed from the free-text portion of the query, so `flooding intitle:jamaica` searches `flooding` across the whole record and `jamaica` in the title only.

| Operator | Restricts the match to | Repeating it |
|---|---|---|
| `intitle:` | the document title | narrows (AND) |
| `insummary:` | the record's summary and highlighted insights | narrows (AND) |
| `ispublisher:` | the publishing organisation's name, never an author's name | narrows (AND) |
| `doctype:` | the document type, matched exactly | widens (OR) |
| `cidid:` | the record id, matched exactly | widens (OR) |
| `hasconcept:` | climate tags, frame/perspective and analysis type | widens (OR) |
| `geography:` | regions the document covers or discusses | widens (OR) |
| `lang:` | document language ISO code | widens (OR) |

`doctype:`, `cidid:`, `lang:`, `hasconcept:` and `geography:` widen rather than narrow.

A value containing a space must be double-quoted: `doctype:"policy brief"`. Written unquoted, it parses as `doctype:policy` plus the free-text word `brief`, which matches nothing. `doctype:` takes one of eight values: `report`, `statement`, `policy brief`, `guide`, `briefing`, `research paper`, `case study`, `fact sheet`. Four of them contain a space.

An unrecognised prefix is not rejected on `/search`. `publisher:oxfam` is not one of the eight operators above, so the whole token is searched as free text and the response is a normal `200` with plausible but unintended results. There is no error to detect this from on HTML search, so use the eight names exactly as spelled.

The public search API (`/api/v1/search`) accepts 16 URL parameters: `q`, `page`, `page_size`, `response_format`, `sort`, `organization`, `published_from`, `published_to`, `added`, `climate_tags`, `geography`, `geography_discussed`, `frame_perspective`, `analysis_type`, `document_type`, `language`. On `/api/v1/search`, an unknown parameter answers `400 application/problem+json` (type `unknown-parameter`, with a suggestion); on `/search`, it is ignored and the page shows a notice.

The HTML search page (`/search`) accepts 14 URL parameters (the API set without `page_size` and `response_format`): `q`, `page`, `sort`, `organization`, `published_from`, `published_to`, `added`, `climate_tags`, `geography`, `geography_discussed`, `frame_perspective`, `analysis_type`, `document_type`, `language`.

Seven parameters filter base attributes: `organization` for the publishing organisation, `added` for the day a record entered the corpus (`YYYY-MM-DD`), `published_from` and `published_to` for date of publication bounds, `sort` (`relevance`, `published`, `published_asc`), `page` for pagination, and `q` for terms. An `added` value that is not a real calendar date is ignored rather than rejected on `/search`. A `q` longer than 200 characters, or one carrying surrounding whitespace, answers `302` to the trimmed form; a page number past the last page answers `302` to the last page.

Seven parameters filter on classifications and facets: `climate_tags`, `geography`, `geography_discussed`, `frame_perspective`, `analysis_type`, `document_type`, `language`. Each is repeatable. Repeating a parameter widens (OR) for all of these: `climate_tags`, `geography`, `geography_discussed`, `frame_perspective`, `analysis_type`, `document_type`, and `language`. Values are matched after lowercasing.

## Core

- [Home](https://climateinsightsdirectory.org/): Entry point, with search, suggested topics, recently indexed documents and corpus statistics.
- [Search](https://climateinsightsdirectory.org/search): Full-text search across every document record. This is the intended way to find documents. It accepts `q` for terms, eight field operators written inside `q`, and 11 filter parameters, all described under Search syntax above. The `added` filter reads the day a record entered the corpus, which is its ingestion date rather than the document's publication date.
- [About Us](https://climateinsightsdirectory.org/about-us): The mission, what the directory does and does not index, the information integrity criteria, how organisations can request removal, who builds the directory, and what each release tier permits.
- [Updates](https://climateinsightsdirectory.org/updates): Running log of content releases, blog posts and guest research: what changed in the corpus and when, including the editorial practice behind each release. Filterable by `?type=content-release`, `?type=blog-post` or `?type=guest-research`.

## Optional

- [Sitemap](https://climateinsightsdirectory.org/sitemap.xml): Machine-readable index of the home page, the updates log and each of its posts, the institutional pages, and document URLs published in per-publisher partitions.
- [Terms of Use](https://climateinsightsdirectory.org/terms-of-use): Conditions for reading, searching and programmatically accessing the directory.
- [Privacy Policy](https://climateinsightsdirectory.org/privacy-policy): What the directory collects, and what it does not.
- [API Documentation](https://climateinsightsdirectory.org/api-docs): Documentation for the public search API, including parameter schemas, response formats and error definitions.
- [OpenAPI Specification](https://climateinsightsdirectory.org/api/v1/openapi.json): OpenAPI 3.1 specification for the public search API.
- [Model Context Protocol (MCP)](https://climateinsightsdirectory.org/api/mcp?utm_source=docs&utm_medium=mcp): Streamable HTTP MCP endpoint providing the `search` and `fetch` tools for AI agents. Add it to Claude Code with `claude mcp add --transport http climate-insights-directory "https://climateinsightsdirectory.org/api/mcp?utm_source=docs&utm_medium=mcp"`.

## Corrections

The directory is a work in progress and welcomes correction. Improper inclusions, errors in a record, recommendations of civil society organisations producing evidence-based insights, and research collaborations all go to info@climateinsightsdirectory.org.