# Deloitte API

> JSON API to search and read the pages, hubs and documents Deloitte publishes at https://deloitte.relayto.com/. No API key and no sign-up, and nothing you call changes the site.

This API is served by RELAYTO, the platform Deloitte publishes this site on.

- Base URL: `https://deloitte.relayto.com/api/site/v1`
- Sandbox base URL: `https://deloitte.relayto.com/api/site/sandbox/v1` (fixed sample data, see [Sandbox and test mode](#sandbox))
- OpenAPI 3.1 description: <https://deloitte.relayto.com/openapi.json> (YAML: <https://deloitte.relayto.com/api/openapi.yaml>)
- API catalog (RFC 9727): <https://deloitte.relayto.com/.well-known/api-catalog>
- This page as Markdown: <https://deloitte.relayto.com/developers.md>; its index for agents: <https://deloitte.relayto.com/developers/llms.txt>
- Index of the whole site for agents: <https://deloitte.relayto.com/llms.txt>

## Quick start

```bash
curl "https://deloitte.relayto.com/api/site/v1/search?q=deloitte"
```

Every answer is JSON. Each result has an `apiUrl` to read it through the API, a `markdownUrl` with its text as Markdown, and a `url` to open it in a browser.

## Authentication

None. There are no API keys, tokens or accounts: send requests without credentials. The API returns only what any visitor of https://deloitte.relayto.com/ can open without signing in. Content behind a password, a sign-in or an email form is never returned in full: a request for it answers `403` with code `not_public`, or, for an email form, returns its description only.

## Sandbox and test mode {#sandbox}

The sandbox answers every operation at `https://deloitte.relayto.com/api/site/sandbox/v1`, over fixed sample data of an imaginary company, Example Industries, instead of this site's content. Its answers never change, so you can test an integration against known results, including the edge cases: `board-pack-2025` answers `403 not_public`, `capital-markets-day-2025` is behind an email form (`access: email-gated`, description only), and the `annual-reports` hub leaves its private item out.

```bash
curl "https://deloitte.relayto.com/api/site/sandbox/v1/contents/annual-report-2025"
curl -X POST "https://deloitte.relayto.com/api/site/sandbox/v1/exports" -H "Content-Type: application/json" -H "Idempotency-Key: test-1" -d '{"hub": "annual-reports"}'
```

The [sandbox guide](https://deloitte.relayto.com/docs/sandbox) lists every sample record. The live API is safe to test against as well: nothing you call changes the site. Every operation reads, except `createExport`, which only starts a job that collects public text. There is no test key to request.

## Endpoints

Paths are relative to the base URL `https://deloitte.relayto.com/api/site/v1` (or the sandbox's). Every operation is a GET, and HEAD works too, except `createExport`, a POST.

| Operation | Request | Returns |
| --- | --- | --- |
| [`getSite`](#get-site) | `GET /` | Get the site and its API links |
| [`searchSite`](#search-site) | `GET /search` | Search the site by keyword |
| [`listPages`](#list-pages) | `GET /pages` | List the pages of the site |
| [`listHubs`](#list-hubs) | `GET /hubs` | List the hubs of the site |
| [`getHub`](#get-hub) | `GET /hubs/{hubSlug}` | Get a hub |
| [`listHubItems`](#list-hub-items) | `GET /hubs/{hubSlug}/items` | List the items in a hub |
| [`listContents`](#list-contents) | `GET /contents` | List the documents of the site |
| [`getContent`](#get-content) | `GET /contents/{contentSlug}` | Get a document or page with its text |
| [`createExport`](#create-export) | `POST /exports` | Start an export of several documents |
| [`getExport`](#get-export) | `GET /exports/{exportId}` | Get an export job and its result |

### Get the site and its API links {#get-site}

`GET /api/site/v1` (operationId `getSite`)

Returns the name and description of this site and the links to its OpenAPI document, developer documentation and site index for agents (llms.txt). Call it first to check the API answers.

Example:

```bash
curl "https://deloitte.relayto.com/api/site/v1"
```

Returns `Site` with `200` (see the OpenAPI description for every field). Errors: `404`, `429`, `500`.

### Search the site by keyword {#search-site}

`GET /api/site/v1/search` (operationId `searchSite`)

Finds pages, hubs and documents of the site whose title or summary contain the words of `q`, best matches first (a word in the title counts more than one in the summary). Each result has a `kind`; read a page or document with `getContent` and a hub with `getHub`.

| Parameter | In | Required | Type | Description |
| --- | --- | --- | --- | --- |
| `q` | query | yes | string | Words to look for in titles and summaries, for example a topic or a document name. |
| `kind` | query | no | string: `page`, `hub`, `content` | Only return results of this kind: `page` (a page of the site), `hub` or `content` (a document). |
| `limit` | query | no | integer, 1-50 | How many items to return, 1-50. |
| `cursor` | query | no | string | Opaque position in the list: `pagination.nextCursor` of the previous page. Omit it for the first page. |

Example:

```bash
curl "https://deloitte.relayto.com/api/site/v1/search?q=deloitte&limit=5"
```

Returns `SearchResults` with `200` (see the OpenAPI description for every field). Errors: `400`, `404`, `429`, `500`.

### List the pages of the site {#list-pages}

`GET /api/site/v1/pages` (operationId `listPages`)

Lists the pages of the website the workspace publishes at its domain (home page first, then the other pages in site order). Empty when the workspace publishes no custom site. Read one with `getContent`.

| Parameter | In | Required | Type | Description |
| --- | --- | --- | --- | --- |
| `limit` | query | no | integer, 1-50 | How many items to return, 1-50. |
| `cursor` | query | no | string | Opaque position in the list: `pagination.nextCursor` of the previous page. Omit it for the first page. |

Example:

```bash
curl "https://deloitte.relayto.com/api/site/v1/pages?limit=5"
```

Returns `ItemList` with `200` (see the OpenAPI description for every field). Errors: `400`, `404`, `429`, `500`.

### List the hubs of the site {#list-hubs}

`GET /api/site/v1/hubs` (operationId `listHubs`)

Lists the public hubs (curated collections of documents) of the site, most recently updated first. List what one contains with `listHubItems`.

| Parameter | In | Required | Type | Description |
| --- | --- | --- | --- | --- |
| `limit` | query | no | integer, 1-50 | How many items to return, 1-50. |
| `cursor` | query | no | string | Opaque position in the list: `pagination.nextCursor` of the previous page. Omit it for the first page. |

Example:

```bash
curl "https://deloitte.relayto.com/api/site/v1/hubs?limit=5"
```

Returns `ItemList` with `200` (see the OpenAPI description for every field). Errors: `400`, `404`, `429`, `500`.

### Get a hub {#get-hub}

`GET /api/site/v1/hubs/{hubSlug}` (operationId `getHub`)

Returns one hub: its title, description, addresses and how many items it holds. A hub behind a password, sign-in or email form answers 403.

| Parameter | In | Required | Type | Description |
| --- | --- | --- | --- | --- |
| `hubSlug` | path | yes | string | The hub's slug or 13-character code, from `listHubs` or `searchSite`. |

Example:

```bash
curl "https://deloitte.relayto.com/api/site/v1/hubs/deloitte-57847acdc9196"
```

Returns `Hub` with `200` (see the OpenAPI description for every field). Errors: `403`, `404`, `429`, `500`.

### List the items in a hub {#list-hub-items}

`GET /api/site/v1/hubs/{hubSlug}/items` (operationId `listHubItems`)

Lists the documents and sub-hubs a hub contains, in the order the hub sets. Items behind a password, sign-in or email form are left out.

| Parameter | In | Required | Type | Description |
| --- | --- | --- | --- | --- |
| `hubSlug` | path | yes | string | The hub's slug or 13-character code, from `listHubs` or `searchSite`. |
| `limit` | query | no | integer, 1-50 | How many items to return, 1-50. |
| `cursor` | query | no | string | Opaque position in the list: `pagination.nextCursor` of the previous page. Omit it for the first page. |

Example:

```bash
curl "https://deloitte.relayto.com/api/site/v1/hubs/deloitte-57847acdc9196/items?limit=5"
```

Returns `ItemList` with `200` (see the OpenAPI description for every field). Errors: `400`, `403`, `404`, `429`, `500`.

### List the documents of the site {#list-contents}

`GET /api/site/v1/contents` (operationId `listContents`)

Lists the public documents, presentations, videos and web pages in the workspace library, most recently updated first. Pages of the site itself are listed by `listPages`. Read one with `getContent`.

| Parameter | In | Required | Type | Description |
| --- | --- | --- | --- | --- |
| `limit` | query | no | integer, 1-50 | How many items to return, 1-50. |
| `cursor` | query | no | string | Opaque position in the list: `pagination.nextCursor` of the previous page. Omit it for the first page. |

Example:

```bash
curl "https://deloitte.relayto.com/api/site/v1/contents?limit=5"
```

Returns `ItemList` with `200` (see the OpenAPI description for every field). Errors: `400`, `404`, `429`, `500`.

### Get a document or page with its text {#get-content}

`GET /api/site/v1/contents/{contentSlug}` (operationId `getContent`)

Returns one document or site page: its metadata and its Markdown version with the full text of every page. When the content asks visitors for an email address or to confirm a notice first, `access` says so and `markdown` holds only its description. A content that is private or behind a password or sign-in answers 403.

| Parameter | In | Required | Type | Description |
| --- | --- | --- | --- | --- |
| `contentSlug` | path | yes | string | The slug or 13-character code of a document or page, from `listContents`, `listPages`, `listHubItems` or `searchSite`. |

Example:

```bash
curl "https://deloitte.relayto.com/api/site/v1/contents/monitor-deloitte-final-report-dfjz70sqtnhl5"
```

Returns `Content` with `200` (see the OpenAPI description for every field). Errors: `403`, `404`, `429`, `500`.

### Start an export of several documents {#create-export}

`POST /api/site/v1/exports` (operationId `createExport`)

Accepts an export of the Markdown of up to 10 documents, given by slug, or of the first 10 documents in a hub. Answers `202 Accepted` at once with the job and its `Location`; `getExport` at that Location returns `items`, read when you ask, so each has the text `getContent` returns at that moment: gated documents come back as `not_public` or with their description only. Nothing is stored and nothing changes: the export id carries the request, so the same body always gets the same id. Send an `Idempotency-Key` header to retry safely; a retry with the same key and body gets the same export and `Idempotent-Replayed: true`. A key reused with a different body is simply a different request and gets a different export.

| Parameter | In | Required | Type | Description |
| --- | --- | --- | --- | --- |
| `Idempotency-Key` | header | no | string | A key for safe retries (a UUID, for example). The export id is derived from the request, so a retry with the same key and body gets the same export, with `Idempotent-Replayed: true`. The API keeps no state, so it cannot tell a key reused with a different body: that is a different request and gets a different export. |

Example:

```bash
curl -i -X POST "https://deloitte.relayto.com/api/site/v1/exports" -H "Content-Type: application/json" -H "Idempotency-Key: $(uuidgen)" -d '{"contents": ["monitor-deloitte-final-report-dfjz70sqtnhl5"]}'
```

Returns `ExportJob` with `202` (see the OpenAPI description for every field). Errors: `400`, `403`, `404`, `429`, `500`.

### Get an export job and its result {#get-export}

`GET /api/site/v1/exports/{exportId}` (operationId `getExport`)

Returns the export, `completed` with `items` read now from the current content (or `failed` with `error` when its hub is gone or no longer public). Each document read counts against the rate limit, as it would through `getContent` (a hub export counts 10). An id this site did not issue, or issued before its domain moved to another workspace, answers 404.

| Parameter | In | Required | Type | Description |
| --- | --- | --- | --- | --- |
| `exportId` | path | yes | string | The `id` createExport returned: a signed token that carries the request. |

Returns `ExportJob` with `200` (see the OpenAPI description for every field). Errors: `404`, `429`, `500`.

## Exports (asynchronous jobs) {#exports}

`createExport` exports the Markdown of up to 10 documents, or of the first 10 documents in a hub. It answers `202 Accepted` at once, with the job (`status: accepted`) and a `Location` header. Read the `Location` (`getExport`) right away: it answers `completed` with `items`, each with the text `getContent` gives at that moment, or a `status` of `not_public` or `not_found` (`failed` when the hub is gone or no longer public).

Nothing is stored and nothing runs in the background: the export id is a signed token that carries the request, and the result is read from the current content each time you ask for it, so a document gated after the export comes back without its text. Export ids do not expire, and they only work on the site, and the live API or sandbox, that issued them. Reading an export counts one request per document against the rate limit.

Retries are safe: the same body always gets the same export id. Send an `Idempotency-Key` header (a UUID, for example) and the answer says `Idempotent-Replayed: true`. Since nothing is stored, a key reused with a different body cannot be detected: that is a different request and gets a different export.

```http
HTTP/1.1 202 Accepted
Location: https://deloitte.relayto.com/api/site/v1/exports/<id>

{"id": "<id>", "status": "accepted", "statusUrl": "https://deloitte.relayto.com/api/site/v1/exports/<id>", "items": null, ...}
```

## Ask in natural language (NLWeb)

The site answers questions with the NLWeb protocol (v0.55) at `https://deloitte.relayto.com/ask`, backed by the same search as `searchSite`. Results are schema.org objects with a name, description and URL.

```bash
curl -X POST "https://deloitte.relayto.com/ask" -H "Content-Type: application/json" -d '{"query": {"text": "deloitte"}}'
```

Send `"prefer": {"streaming": true}` (or `Accept: text/event-stream`) for Server-Sent Events: `start`, one `result` per item, then `complete`. `GET https://deloitte.relayto.com/ask?query=...` works too. No results is an NLWeb failure with code `NO_RESULTS`.

## Pagination

Lists and search return one page at a time: `limit` sets its size (1-50, default 10). The answer's `pagination.nextCursor` is the `cursor` for the next page, and `pagination.hasMore` says whether there is one; `links.next` and a `Link: <...>; rel="next"` header give the full URL. Cursors are opaque: send them back unchanged, and only to the list that issued them (another list answers `400` with code `invalid_cursor`).

```json
{"data": [], "pagination": {"limit": 10, "hasMore": true, "nextCursor": "eyJvIjoxMH0"}, "links": {"self": "...", "next": "..."}}
```

## Errors {#errors}

Every 4xx and 5xx answer is an RFC 9457 problem with `Content-Type: application/problem+json`. `code` is stable and machine-readable, `detail` explains this request, and `hint` says what to do. `type` links to the code's entry below.

```json
{"type": "https://deloitte.relayto.com/developers#error-not_found", "title": "Not found", "status": 404, "detail": "There is no content \"x\" on this site.", "instance": "/api/site/v1/contents/x", "code": "not_found", "hint": "Check the slug with searchSite, listContents or listHubs."}
```

### `invalid_parameter` (400 Invalid parameter) {#error-invalid_parameter}

A query parameter is missing, empty or out of range. Fix the parameter named in `detail` and send the request again.

### `invalid_cursor` (400 Invalid cursor) {#error-invalid_cursor}

The cursor was not issued for this list or query. Send `nextCursor` exactly as the previous page of the same list returned it, or drop it to start over.

### `not_public` (403 Not public) {#error-not_public}

The content or hub exists but is private, or needs a password, a sign-in or an email address to open. Do not retry: no key unlocks it. Send a person to the `url` member to open it in a browser.

### `not_found` (404 Not found) {#error-not_found}

No content, hub or endpoint matches the request on this site. Check the slug with `searchSite`, `listContents` or `listHubs`.

### `invalid_body` (400 Invalid body) {#error-invalid_body}

The JSON body of a POST is missing, malformed or does not match its schema. Fix the body as `detail` says and send it again.

### `method_not_allowed` (405 Method not allowed) {#error-method_not_allowed}

The operation does not accept this HTTP method: everything is GET or HEAD, except POST on /exports and /ask. Use the method the OpenAPI description gives.

### `rate_limited` (429 Too many requests) {#error-rate_limited}

The client sent more requests than the rate limit allows in the current window. Wait for the number of seconds in `Retry-After`, then retry.

### `internal_error` (500 Internal error) {#error-internal_error}

The server failed to answer. Retry with exponential backoff; the request is safe to repeat.

## Rate limits

120 requests per 60 seconds per client IP address. Every answer says how much is left, with the IETF RateLimit headers and their older separate forms:

```http
RateLimit-Policy: "per-client";q=120;w=60
RateLimit: "per-client";r=119;t=60
RateLimit-Limit: 120
RateLimit-Remaining: 119
RateLimit-Reset: 60
```

Past the limit, the API answers `429` with code `rate_limited` and a `Retry-After` header (seconds). Slow down when `RateLimit-Remaining` nears 0.

## Versioning and deprecation {#deprecation}

- The major version is in the path: `/api/site/v1`. The OpenAPI document's `info.version` gives the exact version.
- Within v1, changes are additive only: new endpoints, optional parameters and response fields. Ignore fields you do not know.
- A change that could break a client ships as a new major version (`/api/site/v2`), alongside v1.
- A version being retired keeps working for at least 6 months after it is deprecated. Meanwhile its answers carry a `Deprecation` header (RFC 9745), a `Sunset` header with the retirement date (RFC 8594) and a `Link` with `rel="deprecation"` to this section, which will name the replacement.
- v1 is current and not deprecated.

## Changing the site's content

This API only reads. Deloitte's editors manage the site in RELAYTO, the platform it is published on, including through RELAYTO's MCP server for signed-in workspace members at <https://mcp.relayto.com/mcp> (sign in with a RELAYTO account over OAuth). Neither is needed to read this site. Agents that speak MCP can read the same public content through the site's read-only content server at <https://deloitte.relayto.com/mcp>, with no sign-in either.
