---
title: "Brave AI Mode"
method: POST
path: "/v1/brave/ai-mode"
tags: ["Brave"]
---

# Brave AI Mode

`POST /v1/brave/ai-mode`

## Request

- Base URL: `https://api.prismcrawl.com`
- URL: `https://api.prismcrawl.com/v1/brave/ai-mode`
- Auth: API key in header `x-api-key`

Retrieve a Brave AI Mode answer and its sources. Send a new query with next_page_token to continue the conversation, keeping localization settings unchanged. Public proof and AI streaming can require up to 40 seconds on CPU-constrained runtimes.

## Request body

- CatalogBraveAiModeRequest: Retrieve a Brave AI Mode answer and its sources. Send a new query with next_page_token to continue the conversation, keeping localization settings unchanged. Public proof and AI streaming can require up to 40 seconds on CPU-constrained runtimes.
  - `query` string, required: Text to search for on the selected service. At most 8,192 UTF-8 bytes.
  - `device` 'desktop' | 'mobile' | 'tablet': Device presentation for the source request.
  - `html` boolean: Return the encoded original response rather than parsed JSON.
  - `zero_trace` boolean: Do not archive source or parsed response artifacts.
  - `country` 'ar' | 'au' | 'at' | 'be' | 'br' | 'ca' | 'cl' | 'cn' | 'dk' | 'fi' | 'fr' | 'de' | 'gr' | 'hk' | 'in' | 'id' | 'it' | 'jp' | 'kr' | 'my' | 'mx' | 'nl' | 'nz' | 'no' | 'pl' | 'pt' | 'ph' | 'ru' | 'sa' | 'sg' | 'za' | 'es' | 'se' | 'ch' | 'tw' | 'tr' | 'gb' | 'us': Two-letter country or storefront code, such as us or gb.
  - `language` 'az' | 'bg' | 'br' | 'ca' | 'cs' | 'cy' | 'da' | 'de' | 'el' | 'en' | 'es' | 'et' | 'eu' | 'fi' | 'fr' | 'gl' | 'hr' | 'hu' | 'id' | 'it' | 'ja' | 'ka' | 'ko' | 'lt' | 'lv' | 'ms' | 'nb' | 'nl' | 'pl' | 'pt' | 'ro' | 'ru' | 'sk' | 'sl' | 'sq' | 'sr' | 'sv' | 'sw' | 'th' | 'tr' | 'uk' | 'vi' | 'zh': Provider-supported language code, such as en.
  - `coordinates` object: Latitude and longitude for location targeting.
    - `latitude` number, required
    - `longitude` number, required
  - `safe` 'off' | 'moderate' | 'strict': Provider safe-search filtering.
  - `units_of_measurement` 'metric' | 'us': Metric or US customary units in the answer.
  - `next_page_token` string: Copy the opaque next_page_token from the previous response using the same account, search parameters, device, and result limit. Tokens expire after one hour; restart pagination if rejected or expired.
  - `location` string: Named geographic search origin; mutually exclusive with coordinates. Resend the same name when continuing a conversation.
  - `output` 'json' | 'html' | 'md' | 'null', nullable: Response content format. md returns Markdown in data.content with data.format=md; the JSON success/request envelope remains. Accept: text/markdown also selects Markdown unless output or html=true is explicit. html is an alias for html=true and retains the existing Brotli/Base64 source format.
  - `json_restrictor` string, nullable: Select fields within data.content for JSON or Markdown. Supports comma-separated paths, non-negative array indices, half-open slices [0:3] (either bound may be omitted), [] and .{nested,subqueries}. Up to128 selections and16 levels. Missing fields are omitted; selected declared fields retain their zero defaults. The request envelope and credit headers remain. Include next_page_token when pagination is needed. Incompatible with HTML output. This changes response presentation only; tracing-enabled history retains the complete JSON result.
  - `no_cache` boolean, nullable: Accepted for compatibility. Searches are always fetched without a result cache for either value; normal successful-request billing applies.

## Response `200`

Request completed. One credit is charged, including recognized empty or unavailable source results and source inventory changes discovered after a valid continuation is fetched. These completed responses may preserve verified partial results; operational failures remain errors.

- CatalogBraveAiModeResponse
  - `success` true, required
  - `request_id` string, uuid, required
  - `data` union, required
    - MarkdownSearchData
      - `format` union, required
        - ''
        - 'md'
      - `content` string, required: Markdown derived from the parsed result, after any json_restrictor selection.
    - ProjectedSearchData
      - `format` union, required
        - ''
        - 'json'
      - `content` object, required: Only fields selected by json_restrictor. Selected fields retain their declared types and zero defaults.
    - object
      - `format` 'json', required
      - `content` object, required
        - `has_next_page` boolean, required
        - `next_page_token` string, required: Opaque continuation token valid for one hour and the same account. Retain search parameters, device, and result limit. Restart pagination if the token is expired or rejected; tokens from before the authenticated-token rollout cannot be reused.
        - `results` object[], required
          - `address` union, required
            - string
            - object
            - string[]
          - `answer` string, required
          - `category` string, required
          - `coordinates` object, required
            - `latitude` number, required
            - `longitude` number, required
          - `description` string, required
          - `followup_questions` string[], required
          - `id` string, required
          - `latitude` number, required
          - `longitude` number, required
          - `name` string, required
          - `phone` string, required
          - `position` integer, required
          - `rank` integer, required
          - `rating` union, required
            - number
            - string
          - `rating_max` number, required
          - `rating_source` string, required
          - `review_count` union, required
            - number
            - string
          - `snippet` string, required
          - `sources` object[], required
            - `title` string, required
            - `url` string, required
          - `thumbnail` string, required
          - `title` string, required
          - `type` string, required
          - `url` string, required
          - `website` string, required
        - `search_parameters` object, required
          - `coordinates` object, required
            - `latitude` number, required
            - `longitude` number, required
          - `country` string, required
          - `device` string, required
          - `engine` string, required
          - `language` string, required
          - `location` string, required
          - `operation` string, required
          - `pagination_limited` boolean, required: True when further pagination is unavailable. A completed source-stale request can contain verified results or an empty array; next_page_token is an empty string.
          - `pagination_reason` string, required: stale_cursor means a valid continuation was fetched but its source inventory changed. The HTTP 200 response is billable; keep any verified results and restart pagination.
          - `query` string, required
          - `safe` string, required
          - `type` string, required
          - `units_of_measurement` string, required
        - `serp_features` object[], required
          - `confidence` number, required
          - `engine` string, required
          - `extracted_at` string, required
          - `id` string, required
          - `items` object[], required
            - `link` string, required
            - `product_id` string, required
            - `text` string, required
            - `thumbnail` string, required
            - `title` string, required
          - `links` object[], required
            - `title` string, required
            - `url` string, required
          - `position` object, required
            - `absolute` integer, required
          - `source_result_ids` string[], required
          - `text` string, required
          - `title` string, required
          - `type` string, required
    - GoogleHtmlSearchData
      - `format` union, required
        - ''
        - 'html+br+base64'
      - `content` string, byte, required: Base64-encoded bytes. Decode Base64 first, then decompress the result with Brotli. Available deferred or lazy-loaded thumbnail sources are placed in image src attributes, within size limits; image and video files are not downloaded. When rewrite_links is true, already-resolved opaque Google anchor links use their destination URLs. Unresolved links are preserved.

## Other responses

- `400`: The JSON body, source selection, or continuation is invalid. Locally malformed, forged, expired, or mismatched tokens are rejected before source work and require restarting pagination. A source may also reject an otherwise well-formed selector after acquisition. These errors do not spend a search credit. Source inventory changes discovered during a valid request return a completed response instead.
- `401`: The API token is missing, invalid, expired, or revoked.
- `402`: No credits remain.
- `429`: Account rate limit exceeded.
- `500`: The request could not be completed.

---

[API](https://skmtc.dev/prismcrawl/apis/prismcrawl-api.md) · [All operations](https://skmtc.dev/prismcrawl/apis/prismcrawl-api/llms.txt) · [OpenAPI document](https://skmtc.dev/prismcrawl/apis/prismcrawl-api/revisions/eef17f6c02a5?raw)