diff --git a/.changeset/first-version.md b/.changeset/first-version.md
index 7a6f7b1..b948586 100644
--- a/.changeset/first-version.md
+++ b/.changeset/first-version.md
@@ -7,6 +7,7 @@ First version of `meili`, the Meilisearch CLI built for coding agents.
 - Commands derive from the routes of the Meilisearch API and of the Management API, verb first and resource plural: `meili get health`, `meili get indexes`, `meili get indexes --id movies`.
 - Documents: `meili get documents --index <index>`, narrowed by `--ids` or `--filter`; `meili add documents --index <index> --file <path>` and `meili update documents --index <index> --file <path>`, which read the documents from a file, from stdin with `--file -` or from `--data`; `meili delete documents --index <index> (--ids <ids> | --filter <filter> | --all) --yes`.
 - Indexes: `meili get indexes [--id <index>]`; `meili add indexes --id <index> [--primary-key <field>]`, `meili update indexes --id <index> [--primary-key <field>] [--new-id <index>]`, which renames the index, and `meili delete indexes --id <index> --yes`.
+- Search: `meili search --index <index> --q <query>`, on `POST /indexes/{index_uid}/search`, with every non-experimental search parameter of the route as a flag, from `--filter` and `--sort` to `--hybrid` and `--vector`; `--data` or `--file` carries a whole search body, which a flag of the same field overrides.
 - `meili schema` prints the index of the command tree as JSON, one entry per command with its routes, and `meili schema --command "get indexes"` prints one command in full, with its key actions and whether it returns a task.
 - Connection through `--project-url` and `--project-api-key`, or `MEILI_PROJECT_URL` and `MEILI_PROJECT_API_KEY`; `MEILI_MANAGEMENT_TOKEN` and `MEILI_MANAGEMENT_URL` for the Management API.
 - Several projects in a config file, `~/.config/meili/config.json` or `MEILI_CONFIG`, chosen with `--project` on the command line and never deduced; `meili config list`, `config set` and `config unset` manage the file.
diff --git a/docs/contract/commands.md b/docs/contract/commands.md
index 8d46018..5e6f3c2 100644
--- a/docs/contract/commands.md
+++ b/docs/contract/commands.md
@@ -28,7 +28,7 @@ Every API command declares its routes, `static routes = ["GET /indexes", "GET /i
 7. **Values.** A scalar is passed as it is. A boolean is a bare flag, `--retrieve-vectors`. An array of scalars is comma-separated, `--fields id,title`, as the API's own GET syntax. An object or a nested array is JSON text, `--typo-tolerance '{"enabled":false}'`. The value `null` sends a JSON null, `--expires-at null`, `--stop-words null`.
 8. **Body input.** On every route with a body, `--file <path>` or `--data '<json>'` replaces the body fields; `--file -` reads stdin. Path and query flags stay valid next to them. A body-field flag given together with `--file` or `--data` overrides that field, as a flag overrides its environment variable. When a route's body is not an object of named fields but the value itself, the array of identifiers of `delete documents --ids` or the documents of `add documents`, the body is that value, and its shape selects the route among the ones the command covers: `delete documents --data '[1,2]'` deletes by identifiers, `--data '{"filter":…}'` deletes by filter. Documents are sent as JSON, the only content type the document declares.
 9. **Routes that are not implemented.** The `GET` twin of a `POST` route that takes the same parameters in a body (search, similar, the list of documents); the unit route of a resource whose list filters by identifiers (documents, tasks, batches); the settings sub-routes, since `update settings` sends any subset and `null` resets one setting; `GET /api/v2/meilisearch-versions`. Each is recorded without command in the exceptions.
-10. **Experimental routes** are out of the contract until the feature is stable. Implementing one is a contract change: an exception row with the reason. `compact` is the one exception today.
+10. **Experimental routes** are out of the contract until the feature is stable. Implementing one is a contract change: an exception row with the reason. `compact` is the one exception today. A body field or a query parameter that an experimental feature gates is out of the contract the same way: the command declares no flag for it, and `--data` or `--file` still carries it to the API. `useNetwork` on `get documents`, and `media`, `personalize`, `useNetwork` and `inlineRule` on `search`. These stay out of the exceptions table, which lists routes.
 11. **Destructive commands.** A command of the verb `delete` takes `--yes` and refuses to run without it: `cli_usage`, exit 2, before any network call. `--yes` is a flag to write, never a question the CLI asks, see `output.md`.
 
 | Routes | Command |
diff --git a/oclif.manifest.json b/oclif.manifest.json
index b4d405f..208e8b1 100644
--- a/oclif.manifest.json
+++ b/oclif.manifest.json
@@ -2705,6 +2705,410 @@
         "schema.js"
       ]
     },
+    "search": {
+      "aliases": [],
+      "args": {},
+      "description": "Search the documents of an index, with --q and the search parameters of the route.",
+      "examples": [
+        "<%= config.bin %> <%= command.id %> --index movies --q nosferatu",
+        "<%= config.bin %> <%= command.id %> --index movies --q nosferatu --output json",
+        "<%= config.bin %> <%= command.id %> --index movies --q horror --filter 'rating > 4' --sort rating:desc --limit 5",
+        "<%= config.bin %> <%= command.id %> --index movies --facets genres --attributes-to-retrieve id,title",
+        "<%= config.bin %> <%= command.id %> --index movies --data '{\"q\":\"nosferatu\",\"hybrid\":{\"embedder\":\"default\",\"semanticRatio\":0.5}}'"
+      ],
+      "flags": {
+        "output": {
+          "char": "o",
+          "description": "json (default): one document, the API body as it is, indented in a terminal, compact when piped. ndjson: one item per line, list commands only, streams with --all.",
+          "name": "output",
+          "hasDynamicHelp": false,
+          "helpValue": "<mode>",
+          "multiple": false,
+          "options": [
+            "json",
+            "ndjson"
+          ],
+          "type": "option"
+        },
+        "debug": {
+          "description": "Print every HTTP request and response on stderr, secrets redacted.",
+          "name": "debug",
+          "allowNo": false,
+          "type": "boolean"
+        },
+        "project-url": {
+          "description": "Project URL, the Meilisearch instance to talk to.",
+          "name": "project-url",
+          "hasDynamicHelp": false,
+          "helpValue": "<url>",
+          "multiple": false,
+          "type": "option"
+        },
+        "project-api-key": {
+          "description": "Project API key. `-` reads it from stdin.",
+          "name": "project-api-key",
+          "hasDynamicHelp": false,
+          "helpValue": "<key>",
+          "multiple": false,
+          "type": "option"
+        },
+        "project": {
+          "description": "A project of the config file. Cannot be combined with --project-url. Required when the file has several projects.",
+          "name": "project",
+          "hasDynamicHelp": false,
+          "helpValue": "<name>",
+          "multiple": false,
+          "type": "option"
+        },
+        "index": {
+          "description": "The index to search in.",
+          "name": "index",
+          "required": true,
+          "hasDynamicHelp": false,
+          "helpValue": "<index>",
+          "multiple": false,
+          "type": "option"
+        },
+        "q": {
+          "description": "The search terms. Omit it for a placeholder search, which returns everything.",
+          "name": "q",
+          "hasDynamicHelp": false,
+          "helpValue": "<query>",
+          "multiple": false,
+          "type": "option"
+        },
+        "offset": {
+          "description": "Number of documents to skip. Ignored when --page or --hits-per-page is given.",
+          "name": "offset",
+          "hasDynamicHelp": false,
+          "helpValue": "<n>",
+          "multiple": false,
+          "type": "option"
+        },
+        "limit": {
+          "description": "Maximum number of documents to return, 20 by default. Ignored when --page or --hits-per-page is given.",
+          "name": "limit",
+          "hasDynamicHelp": false,
+          "helpValue": "<n>",
+          "multiple": false,
+          "type": "option"
+        },
+        "page": {
+          "description": "Page to return, 1-indexed. The response then carries totalHits and totalPages instead of estimatedTotalHits.",
+          "name": "page",
+          "hasDynamicHelp": false,
+          "helpValue": "<n>",
+          "multiple": false,
+          "type": "option"
+        },
+        "hits-per-page": {
+          "description": "Number of documents per page, used with --page. 0 counts without returning any.",
+          "name": "hits-per-page",
+          "hasDynamicHelp": false,
+          "helpValue": "<n>",
+          "multiple": false,
+          "type": "option"
+        },
+        "attributes-to-retrieve": {
+          "description": "Document attributes to return, comma-separated. Defaults to the displayed ones.",
+          "name": "attributes-to-retrieve",
+          "hasDynamicHelp": false,
+          "helpValue": "<fields>",
+          "multiple": false,
+          "type": "option"
+        },
+        "attributes-to-crop": {
+          "description": "Attributes to crop in _formatted, comma-separated, each optionally as `attribute:length`.",
+          "name": "attributes-to-crop",
+          "hasDynamicHelp": false,
+          "helpValue": "<fields>",
+          "multiple": false,
+          "type": "option"
+        },
+        "crop-length": {
+          "description": "Maximum number of words in a cropped value, 10 by default.",
+          "name": "crop-length",
+          "hasDynamicHelp": false,
+          "helpValue": "<n>",
+          "multiple": false,
+          "type": "option"
+        },
+        "crop-marker": {
+          "description": "String marking where a cropped value was cut, `…` by default.",
+          "name": "crop-marker",
+          "hasDynamicHelp": false,
+          "helpValue": "<string>",
+          "multiple": false,
+          "type": "option"
+        },
+        "attributes-to-highlight": {
+          "description": "Attributes to highlight in _formatted, comma-separated.",
+          "name": "attributes-to-highlight",
+          "hasDynamicHelp": false,
+          "helpValue": "<fields>",
+          "multiple": false,
+          "type": "option"
+        },
+        "highlight-pre-tag": {
+          "description": "String inserted before each highlighted term, `<em>` by default.",
+          "name": "highlight-pre-tag",
+          "hasDynamicHelp": false,
+          "helpValue": "<string>",
+          "multiple": false,
+          "type": "option"
+        },
+        "highlight-post-tag": {
+          "description": "String inserted after each highlighted term, `</em>` by default.",
+          "name": "highlight-post-tag",
+          "hasDynamicHelp": false,
+          "helpValue": "<string>",
+          "multiple": false,
+          "type": "option"
+        },
+        "show-matches-position": {
+          "description": "Add _matchesPosition to each hit, the byte offset of every matched term.",
+          "name": "show-matches-position",
+          "allowNo": false,
+          "type": "boolean"
+        },
+        "filter": {
+          "description": "Filter expression, for example `rating > 4`. Its array-of-arrays form goes through --data.",
+          "name": "filter",
+          "hasDynamicHelp": false,
+          "helpValue": "<filter>",
+          "multiple": false,
+          "type": "option"
+        },
+        "sort": {
+          "description": "Attributes to sort on, comma-separated, each as `attribute:asc` or `:desc`.",
+          "name": "sort",
+          "hasDynamicHelp": false,
+          "helpValue": "<sort>",
+          "multiple": false,
+          "type": "option"
+        },
+        "distinct": {
+          "description": "Return one document per value of this attribute, which must be filterable.",
+          "name": "distinct",
+          "hasDynamicHelp": false,
+          "helpValue": "<attribute>",
+          "multiple": false,
+          "type": "option"
+        },
+        "facets": {
+          "description": "Attributes to count in facetDistribution, comma-separated. Patterns such as `*` are accepted.",
+          "name": "facets",
+          "hasDynamicHelp": false,
+          "helpValue": "<facets>",
+          "multiple": false,
+          "type": "option"
+        },
+        "matching-strategy": {
+          "description": "How query terms are relaxed when there are not enough results.",
+          "name": "matching-strategy",
+          "hasDynamicHelp": false,
+          "multiple": false,
+          "options": [
+            "last",
+            "all",
+            "frequency"
+          ],
+          "type": "option"
+        },
+        "attributes-to-search-on": {
+          "description": "Restrict the search to these searchable attributes, comma-separated.",
+          "name": "attributes-to-search-on",
+          "hasDynamicHelp": false,
+          "helpValue": "<fields>",
+          "multiple": false,
+          "type": "option"
+        },
+        "ranking-score-threshold": {
+          "description": "Drop the hits whose ranking score is below this value, between 0.0 and 1.0.",
+          "name": "ranking-score-threshold",
+          "hasDynamicHelp": false,
+          "helpValue": "<number>",
+          "multiple": false,
+          "type": "option"
+        },
+        "locales": {
+          "description": "ISO-639 locales of the query, comma-separated, instead of auto-detection.",
+          "name": "locales",
+          "hasDynamicHelp": false,
+          "helpValue": "<locales>",
+          "multiple": false,
+          "type": "option"
+        },
+        "hybrid": {
+          "description": "Hybrid search as JSON, for example `{\"embedder\":\"default\",\"semanticRatio\":0.5}`.",
+          "name": "hybrid",
+          "hasDynamicHelp": false,
+          "helpValue": "<json>",
+          "multiple": false,
+          "type": "option"
+        },
+        "vector": {
+          "description": "Query vector for a user-provided embedder, comma-separated numbers.",
+          "name": "vector",
+          "hasDynamicHelp": false,
+          "helpValue": "<numbers>",
+          "multiple": false,
+          "type": "option"
+        },
+        "retrieve-vectors": {
+          "description": "Add _vectors to each hit, the embeddings of the document.",
+          "name": "retrieve-vectors",
+          "allowNo": false,
+          "type": "boolean"
+        },
+        "show-ranking-score": {
+          "description": "Add _rankingScore to each hit, between 0.0 and 1.0.",
+          "name": "show-ranking-score",
+          "allowNo": false,
+          "type": "boolean"
+        },
+        "show-ranking-score-details": {
+          "description": "Add _rankingScoreDetails to each hit, the score of every ranking rule.",
+          "name": "show-ranking-score-details",
+          "allowNo": false,
+          "type": "boolean"
+        },
+        "show-performance-details": {
+          "description": "Add performanceDetails to the response, a timing breakdown of the query.",
+          "name": "show-performance-details",
+          "allowNo": false,
+          "type": "boolean"
+        },
+        "file": {
+          "description": "Read the body from a file. `-` reads from stdin.",
+          "exclusive": [
+            "data"
+          ],
+          "name": "file",
+          "hasDynamicHelp": false,
+          "helpValue": "<path>",
+          "multiple": false,
+          "type": "option"
+        },
+        "data": {
+          "description": "Inline JSON body.",
+          "name": "data",
+          "hasDynamicHelp": false,
+          "helpValue": "<json>",
+          "multiple": false,
+          "type": "option"
+        }
+      },
+      "hasDynamicHelp": false,
+      "hiddenAliases": [],
+      "id": "search",
+      "pluginAlias": "@meilisearch/cli",
+      "pluginName": "@meilisearch/cli",
+      "pluginType": "core",
+      "strict": true,
+      "enableJsonFlag": false,
+      "api": "meilisearch",
+      "routes": [
+        "POST /indexes/{index_uid}/search"
+      ],
+      "keyActions": [
+        "search"
+      ],
+      "connectionFlags": {
+        "project-url": {
+          "description": "Project URL, the Meilisearch instance to talk to.",
+          "helpValue": "<url>",
+          "input": [],
+          "multiple": false,
+          "type": "option"
+        },
+        "project-api-key": {
+          "description": "Project API key. `-` reads it from stdin.",
+          "helpValue": "<key>",
+          "input": [],
+          "multiple": false,
+          "type": "option"
+        },
+        "project": {
+          "description": "A project of the config file. Cannot be combined with --project-url. Required when the file has several projects.",
+          "helpValue": "<name>",
+          "input": [],
+          "multiple": false,
+          "type": "option"
+        }
+      },
+      "waitFlags": {
+        "wait": {
+          "description": "Poll the task until it reaches a final status, then print the final task.",
+          "allowNo": false,
+          "type": "boolean"
+        },
+        "timeout": {
+          "description": "Maximum wait time, for example 30s, 2m. Default 60s. Only with --wait.",
+          "helpValue": "<duration>",
+          "dependsOn": [
+            "wait"
+          ],
+          "input": [],
+          "multiple": false,
+          "type": "option"
+        }
+      },
+      "paginationFlags": {
+        "limit": {
+          "description": "Passed to the API.",
+          "helpValue": "<n>",
+          "input": [],
+          "multiple": false,
+          "type": "option"
+        },
+        "offset": {
+          "description": "Passed to the API.",
+          "helpValue": "<n>",
+          "input": [],
+          "multiple": false,
+          "type": "option"
+        },
+        "all": {
+          "description": "Fetch every page. Streams with ndjson, buffers into one array with json.",
+          "allowNo": false,
+          "type": "boolean"
+        }
+      },
+      "bodyInputFlags": {
+        "file": {
+          "description": "Read the body from a file. `-` reads from stdin.",
+          "helpValue": "<path>",
+          "exclusive": [
+            "data"
+          ],
+          "input": [],
+          "multiple": false,
+          "type": "option"
+        },
+        "data": {
+          "description": "Inline JSON body.",
+          "helpValue": "<json>",
+          "input": [],
+          "multiple": false,
+          "type": "option"
+        }
+      },
+      "confirmFlags": {
+        "yes": {
+          "required": true,
+          "description": "Confirm the deletion. Nothing is sent to the API without it.",
+          "allowNo": false,
+          "type": "boolean"
+        }
+      },
+      "isESM": true,
+      "relativePath": [
+        "dist",
+        "commands",
+        "search.js"
+      ]
+    },
     "update:documents": {
       "aliases": [],
       "args": {},
diff --git a/src/commands/search.ts b/src/commands/search.ts
new file mode 100644
index 0000000..a684ce9
--- /dev/null
+++ b/src/commands/search.ts
@@ -0,0 +1,215 @@
+import { Flags } from "@oclif/core";
+import type { HybridSearch, SearchParams } from "meilisearch";
+import { type Api, BaseCommand } from "../lib/base-command.js";
+import { readJsonObjectBody } from "../lib/input.js";
+
+export default class Search extends BaseCommand {
+  static api: Api = "meilisearch";
+
+  static routes = ["POST /indexes/{index_uid}/search"];
+
+  static keyActions = ["search"];
+
+  static override description =
+    "Search the documents of an index, with --q and the search parameters of the route.";
+
+  static override examples = [
+    "<%= config.bin %> <%= command.id %> --index movies --q nosferatu",
+    "<%= config.bin %> <%= command.id %> --index movies --q nosferatu --output json",
+    "<%= config.bin %> <%= command.id %> --index movies --q horror --filter 'rating > 4' --sort rating:desc --limit 5",
+    "<%= config.bin %> <%= command.id %> --index movies --facets genres --attributes-to-retrieve id,title",
+    '<%= config.bin %> <%= command.id %> --index movies --data \'{"q":"nosferatu","hybrid":{"embedder":"default","semanticRatio":0.5}}\'',
+  ];
+
+  static override flags = {
+    ...BaseCommand.connectionFlags,
+    index: Flags.string({
+      required: true,
+      description: "The index to search in.",
+      helpValue: "<index>",
+    }),
+    q: Flags.string({
+      description: "The search terms. Omit it for a placeholder search, which returns everything.",
+      helpValue: "<query>",
+    }),
+    offset: Flags.integer({
+      description: "Number of documents to skip. Ignored when --page or --hits-per-page is given.",
+      helpValue: "<n>",
+    }),
+    limit: Flags.integer({
+      description:
+        "Maximum number of documents to return, 20 by default. Ignored when --page or --hits-per-page is given.",
+      helpValue: "<n>",
+    }),
+    page: Flags.integer({
+      description:
+        "Page to return, 1-indexed. The response then carries totalHits and totalPages instead of estimatedTotalHits.",
+      helpValue: "<n>",
+    }),
+    "hits-per-page": Flags.integer({
+      description:
+        "Number of documents per page, used with --page. 0 counts without returning any.",
+      helpValue: "<n>",
+    }),
+    "attributes-to-retrieve": Flags.string({
+      description:
+        "Document attributes to return, comma-separated. Defaults to the displayed ones.",
+      helpValue: "<fields>",
+    }),
+    "attributes-to-crop": Flags.string({
+      description:
+        "Attributes to crop in _formatted, comma-separated, each optionally as `attribute:length`.",
+      helpValue: "<fields>",
+    }),
+    "crop-length": Flags.integer({
+      description: "Maximum number of words in a cropped value, 10 by default.",
+      helpValue: "<n>",
+    }),
+    "crop-marker": Flags.string({
+      description: "String marking where a cropped value was cut, `…` by default.",
+      helpValue: "<string>",
+    }),
+    "attributes-to-highlight": Flags.string({
+      description: "Attributes to highlight in _formatted, comma-separated.",
+      helpValue: "<fields>",
+    }),
+    "highlight-pre-tag": Flags.string({
+      description: "String inserted before each highlighted term, `<em>` by default.",
+      helpValue: "<string>",
+    }),
+    "highlight-post-tag": Flags.string({
+      description: "String inserted after each highlighted term, `</em>` by default.",
+      helpValue: "<string>",
+    }),
+    "show-matches-position": Flags.boolean({
+      description: "Add _matchesPosition to each hit, the byte offset of every matched term.",
+    }),
+    filter: Flags.string({
+      description:
+        "Filter expression, for example `rating > 4`. Its array-of-arrays form goes through --data.",
+      helpValue: "<filter>",
+    }),
+    sort: Flags.string({
+      description: "Attributes to sort on, comma-separated, each as `attribute:asc` or `:desc`.",
+      helpValue: "<sort>",
+    }),
+    distinct: Flags.string({
+      description: "Return one document per value of this attribute, which must be filterable.",
+      helpValue: "<attribute>",
+    }),
+    facets: Flags.string({
+      description:
+        "Attributes to count in facetDistribution, comma-separated. Patterns such as `*` are accepted.",
+      helpValue: "<facets>",
+    }),
+    "matching-strategy": Flags.option({
+      options: ["last", "all", "frequency"] as const,
+      description: "How query terms are relaxed when there are not enough results.",
+    })(),
+    "attributes-to-search-on": Flags.string({
+      description: "Restrict the search to these searchable attributes, comma-separated.",
+      helpValue: "<fields>",
+    }),
+    "ranking-score-threshold": Flags.string({
+      description: "Drop the hits whose ranking score is below this value, between 0.0 and 1.0.",
+      helpValue: "<number>",
+    }),
+    locales: Flags.string({
+      description: "ISO-639 locales of the query, comma-separated, instead of auto-detection.",
+      helpValue: "<locales>",
+    }),
+    hybrid: Flags.string({
+      description:
+        'Hybrid search as JSON, for example `{"embedder":"default","semanticRatio":0.5}`.',
+      helpValue: "<json>",
+    }),
+    vector: Flags.string({
+      description: "Query vector for a user-provided embedder, comma-separated numbers.",
+      helpValue: "<numbers>",
+    }),
+    "retrieve-vectors": Flags.boolean({
+      description: "Add _vectors to each hit, the embeddings of the document.",
+    }),
+    "show-ranking-score": Flags.boolean({
+      description: "Add _rankingScore to each hit, between 0.0 and 1.0.",
+    }),
+    "show-ranking-score-details": Flags.boolean({
+      description: "Add _rankingScoreDetails to each hit, the score of every ranking rule.",
+    }),
+    "show-performance-details": Flags.boolean({
+      description: "Add performanceDetails to the response, a timing breakdown of the query.",
+    }),
+    ...BaseCommand.bodyInputFlags,
+  };
+
+  async run(): Promise<void> {
+    const { flags } = await this.parse(Search);
+    const fromFlags: SearchParams = {
+      q: flags.q,
+      offset: flags.offset,
+      limit: flags.limit,
+      page: flags.page,
+      hitsPerPage: flags["hits-per-page"],
+      attributesToRetrieve: flags["attributes-to-retrieve"]?.split(","),
+      attributesToCrop: flags["attributes-to-crop"]?.split(","),
+      cropLength: flags["crop-length"],
+      cropMarker: flags["crop-marker"],
+      attributesToHighlight: flags["attributes-to-highlight"]?.split(","),
+      highlightPreTag: flags["highlight-pre-tag"],
+      highlightPostTag: flags["highlight-post-tag"],
+      showMatchesPosition: flags["show-matches-position"] ? true : undefined,
+      filter: flags.filter,
+      sort: flags.sort?.split(","),
+      distinct: flags.distinct,
+      facets: flags.facets?.split(","),
+      matchingStrategy: flags["matching-strategy"],
+      attributesToSearchOn: flags["attributes-to-search-on"]?.split(","),
+      rankingScoreThreshold: asNumber(flags["ranking-score-threshold"]) as number | undefined,
+      locales: flags.locales?.split(","),
+      hybrid: asJson(flags.hybrid) as HybridSearch | undefined,
+      vector: flags.vector?.split(",").map(asNumber) as number[] | undefined,
+      retrieveVectors: flags["retrieve-vectors"] ? true : undefined,
+      showRankingScore: flags["show-ranking-score"] ? true : undefined,
+      showRankingScoreDetails: flags["show-ranking-score-details"] ? true : undefined,
+      showPerformanceDetails: flags["show-performance-details"] ? true : undefined,
+    };
+
+    // The body of --file or --data first, then the flags, which override a field of it (rule 8).
+    // The undefined entries are dropped rather than spread: an absent flag would otherwise
+    // erase the field the body carries, since JSON.stringify leaves an undefined value out.
+    const params = {
+      ...readJsonObjectBody(flags),
+      ...Object.fromEntries(Object.entries(fromFlags).filter(([, value]) => value !== undefined)),
+    } as SearchParams;
+
+    await this.output(await this.meilisearch.index(flags.index).search(undefined, params));
+  }
+}
+
+/**
+ * The two conversions a flag needs before it can be a body field (rule 7): the
+ * text of a number, and the JSON text of an object. Text that does not convert
+ * is sent as it is, so that the API names the value it refused, with its own
+ * code, rather than the CLI judging a value the API owns. The CLI only refuses
+ * what it cannot send at all, as `--data` and `--file`, which it has to parse
+ * before it can merge them.
+ */
+function asNumber(value: string | undefined): number | string | undefined {
+  if (value === undefined) return undefined;
+  const text = value.trim();
+  // `Number` alone would not do: it reads `0x1` as 1 and `1e999` as Infinity,
+  // which JSON writes as null. Both turn what the agent typed into another
+  // value, silently. Only text that spells a decimal number becomes one.
+  return DECIMAL.test(text) && Number.isFinite(Number(text)) ? Number(text) : value;
+}
+
+const DECIMAL = /^[+-]?(?:\d+(?:\.\d*)?|\.\d+)(?:[eE][+-]?\d+)?$/;
+
+function asJson(value: string | undefined): unknown {
+  if (value === undefined) return undefined;
+  try {
+    return JSON.parse(value);
+  } catch {
+    return value;
+  }
+}
diff --git a/test/commands/__snapshots__/search.test.ts.snap b/test/commands/__snapshots__/search.test.ts.snap
new file mode 100644
index 0000000..50e66f7
--- /dev/null
+++ b/test/commands/__snapshots__/search.test.ts.snap
@@ -0,0 +1,145 @@
+// Vitest Snapshot v1, https://vitest.dev/guide/snapshot.html
+
+exports[`meili search > passes the API error of a missing index through and exits 4 1`] = `
+{
+  "exitCode": 4,
+  "stderr": "{"error":{"code":"index_not_found","type":"invalid_request","message":"Index \`missing\` not found.","link":"https://docs.meilisearch.com/errors#index_not_found"}}
+",
+  "stdout": "",
+}
+`;
+
+exports[`meili search > prints its help without any network call 1`] = `
+"Search the documents of an index, with --q and the search parameters of the route.
+
+USAGE
+  $ meili search --index <index> [-o <mode>] [--debug]
+    [--project-url <url>] [--project-api-key <key>] [--project <name>] [--q
+    <query>] [--offset <n>] [--limit <n>] [--page <n>] [--hits-per-page <n>]
+    [--attributes-to-retrieve <fields>] [--attributes-to-crop <fields>]
+    [--crop-length <n>] [--crop-marker <string>] [--attributes-to-highlight
+    <fields>] [--highlight-pre-tag <string>] [--highlight-post-tag <string>]
+    [--show-matches-position] [--filter <filter>] [--sort <sort>] [--distinct
+    <attribute>] [--facets <facets>] [--matching-strategy last|all|frequency]
+    [--attributes-to-search-on <fields>] [--ranking-score-threshold <number>]
+    [--locales <locales>] [--hybrid <json>] [--vector <numbers>]
+    [--retrieve-vectors] [--show-ranking-score] [--show-ranking-score-details]
+    [--show-performance-details] [--file <path> | --data <json>]
+
+FLAGS
+  -o, --output=<mode>                     json (default): one document, the API
+                                          body as it is, indented in a terminal,
+                                          compact when piped. ndjson: one item
+                                          per line, list commands only, streams
+                                          with --all.
+      --attributes-to-crop=<fields>       Attributes to crop in _formatted,
+                                          comma-separated, each optionally as
+                                          \`attribute:length\`.
+      --attributes-to-highlight=<fields>  Attributes to highlight in _formatted,
+                                          comma-separated.
+      --attributes-to-retrieve=<fields>   Document attributes to return,
+                                          comma-separated. Defaults to the
+                                          displayed ones.
+      --attributes-to-search-on=<fields>  Restrict the search to these
+                                          searchable attributes,
+                                          comma-separated.
+      --crop-length=<n>                   Maximum number of words in a cropped
+                                          value, 10 by default.
+      --crop-marker=<string>              String marking where a cropped value
+                                          was cut, \`…\` by default.
+      --data=<json>                       Inline JSON body.
+      --debug                             Print every HTTP request and response
+                                          on stderr, secrets redacted.
+      --distinct=<attribute>              Return one document per value of this
+                                          attribute, which must be filterable.
+      --facets=<facets>                   Attributes to count in
+                                          facetDistribution, comma-separated.
+                                          Patterns such as \`*\` are accepted.
+      --file=<path>                       Read the body from a file. \`-\` reads
+                                          from stdin.
+      --filter=<filter>                   Filter expression, for example \`rating
+                                          > 4\`. Its array-of-arrays form goes
+                                          through --data.
+      --highlight-post-tag=<string>       String inserted after each highlighted
+                                          term, \`</em>\` by default.
+      --highlight-pre-tag=<string>        String inserted before each
+                                          highlighted term, \`<em>\` by default.
+      --hits-per-page=<n>                 Number of documents per page, used
+                                          with --page. 0 counts without
+                                          returning any.
+      --hybrid=<json>                     Hybrid search as JSON, for example
+                                          \`{"embedder":"default","semanticRatio"
+                                          :0.5}\`.
+      --index=<index>                     (required) The index to search in.
+      --limit=<n>                         Maximum number of documents to return,
+                                          20 by default. Ignored when --page or
+                                          --hits-per-page is given.
+      --locales=<locales>                 ISO-639 locales of the query,
+                                          comma-separated, instead of
+                                          auto-detection.
+      --matching-strategy=<option>        How query terms are relaxed when there
+                                          are not enough results.
+                                          <options: last|all|frequency>
+      --offset=<n>                        Number of documents to skip. Ignored
+                                          when --page or --hits-per-page is
+                                          given.
+      --page=<n>                          Page to return, 1-indexed. The
+                                          response then carries totalHits and
+                                          totalPages instead of
+                                          estimatedTotalHits.
+      --project=<name>                    A project of the config file. Cannot
+                                          be combined with --project-url.
+                                          Required when the file has several
+                                          projects.
+      --project-api-key=<key>             Project API key. \`-\` reads it from
+                                          stdin.
+      --project-url=<url>                 Project URL, the Meilisearch instance
+                                          to talk to.
+      --q=<query>                         The search terms. Omit it for a
+                                          placeholder search, which returns
+                                          everything.
+      --ranking-score-threshold=<number>  Drop the hits whose ranking score is
+                                          below this value, between 0.0 and 1.0.
+      --retrieve-vectors                  Add _vectors to each hit, the
+                                          embeddings of the document.
+      --show-matches-position             Add _matchesPosition to each hit, the
+                                          byte offset of every matched term.
+      --show-performance-details          Add performanceDetails to the
+                                          response, a timing breakdown of the
+                                          query.
+      --show-ranking-score                Add _rankingScore to each hit, between
+                                          0.0 and 1.0.
+      --show-ranking-score-details        Add _rankingScoreDetails to each hit,
+                                          the score of every ranking rule.
+      --sort=<sort>                       Attributes to sort on,
+                                          comma-separated, each as
+                                          \`attribute:asc\` or \`:desc\`.
+      --vector=<numbers>                  Query vector for a user-provided
+                                          embedder, comma-separated numbers.
+
+DESCRIPTION
+  Search the documents of an index, with --q and the search parameters of the
+  route.
+
+EXAMPLES
+  $ meili search --index movies --q nosferatu
+
+  $ meili search --index movies --q nosferatu --output json
+
+  $ meili search --index movies --q horror --filter 'rating > 4' --sort rating:desc --limit 5
+
+  $ meili search --index movies --facets genres --attributes-to-retrieve id,title
+
+  $ meili search --index movies --data '{"q":"nosferatu","hybrid":{"embedder":"default","semanticRatio":0.5}}'
+
+"
+`;
+
+exports[`meili search > prints the API envelope as it is in JSON 1`] = `
+{
+  "exitCode": 0,
+  "stderr": "",
+  "stdout": "{"hits":[{"id":1,"title":"Nosferatu","rating":4.5},{"id":2,"title":"Nosferatu the Vampyre","rating":4}],"query":"nosferatu","processingTimeMs":1,"limit":20,"offset":0,"estimatedTotalHits":2}
+",
+}
+`;
diff --git a/test/commands/search.test.ts b/test/commands/search.test.ts
new file mode 100644
index 0000000..b5ae9a4
--- /dev/null
+++ b/test/commands/search.test.ts
@@ -0,0 +1,156 @@
+import { describe, expect, it } from "vitest";
+import Search from "../../src/commands/search.js";
+import { MEILISEARCH_URL, meilisearchHandlers } from "../helpers/meilisearch-handlers.js";
+import { recordRequests, server, useMockServer } from "../helpers/msw.js";
+import { runCli, runCommand } from "../helpers/run-command.js";
+
+useMockServer();
+
+const env = { MEILI_PROJECT_URL: MEILISEARCH_URL };
+const hits = [
+  { id: 1, title: "Nosferatu", rating: 4.5 },
+  { id: 2, title: "Nosferatu the Vampyre", rating: 4 },
+];
+
+/** The body the CLI sent to POST /indexes/{uid}/search, request by request. */
+async function bodiesOf(argv: string[]): Promise<unknown[]> {
+  const requests = await recordRequests(async () => {
+    await runCommand(Search, argv, { env });
+  });
+  return requests.map((request) => JSON.parse(request.body));
+}
+
+describe("meili search", () => {
+  it("prints its help without any network call", async () => {
+    const result = await runCli(["search", "--help"]);
+    expect(result.exitCode).toBe(0);
+    expect(result.stderr).toBe("");
+    expect(result.stdout).toMatchSnapshot();
+  });
+
+  it("prints the API envelope as it is in JSON", async () => {
+    server.use(meilisearchHandlers.search(hits));
+    const result = await runCommand(
+      Search,
+      ["--index", "movies", "--q", "nosferatu", "--output", "json"],
+      { env },
+    );
+    expect(result).toMatchSnapshot();
+  });
+
+  it("sends the search parameters in the body of the route", async () => {
+    server.use(meilisearchHandlers.search(hits));
+    const bodies = await bodiesOf([
+      "--index",
+      "movies",
+      "--q",
+      "nosferatu",
+      "--filter",
+      "rating > 4",
+      "--sort",
+      "rating:desc",
+      "--facets",
+      "genres",
+      "--attributes-to-retrieve",
+      "id,title",
+      "--matching-strategy",
+      "all",
+      "--ranking-score-threshold",
+      "0.7",
+      "--vector",
+      "0.1,0.2",
+      "--hybrid",
+      '{"embedder":"default","semanticRatio":0.5}',
+      "--show-ranking-score",
+      "--limit",
+      "2",
+      "--offset",
+      "1",
+    ]);
+    expect(bodies).toEqual([
+      {
+        q: "nosferatu",
+        filter: "rating > 4",
+        sort: ["rating:desc"],
+        facets: ["genres"],
+        attributesToRetrieve: ["id", "title"],
+        matchingStrategy: "all",
+        rankingScoreThreshold: 0.7,
+        vector: [0.1, 0.2],
+        hybrid: { embedder: "default", semanticRatio: 0.5 },
+        showRankingScore: true,
+        limit: 2,
+        offset: 1,
+      },
+    ]);
+  });
+
+  it("sends an empty body for a placeholder search", async () => {
+    server.use(meilisearchHandlers.search(hits));
+    expect(await bodiesOf(["--index", "movies"])).toEqual([{}]);
+  });
+
+  it("takes the body from --data, which a flag of the same field overrides", async () => {
+    server.use(meilisearchHandlers.search(hits));
+    expect(
+      await bodiesOf(["--index", "movies", "--data", '{"q":"nosferatu","filter":[["a = 1"]]}']),
+    ).toEqual([{ q: "nosferatu", filter: [["a = 1"]] }]);
+    expect(
+      await bodiesOf(["--index", "movies", "--data", '{"q":"nosferatu"}', "--q", "alien"]),
+    ).toEqual([{ q: "alien" }]);
+  });
+
+  it("sends a value that does not convert as the text it was given, for the API to refuse", async () => {
+    server.use(meilisearchHandlers.search(hits));
+    expect(
+      await bodiesOf([
+        "--index",
+        "movies",
+        "--vector",
+        "0.1,nope",
+        "--ranking-score-threshold",
+        "abc",
+        "--hybrid",
+        "{nope}",
+      ]),
+    ).toEqual([{ vector: [0.1, "nope"], rankingScoreThreshold: "abc", hybrid: "{nope}" }]);
+  });
+
+  it("converts the text that spells a decimal number, and only that text", async () => {
+    server.use(meilisearchHandlers.search(hits));
+    for (const [text, sent] of [
+      ["0.5", 0.5],
+      [".5", 0.5],
+      ["-1e-3", -0.001],
+      // `Number` would read these as 1 and Infinity, another value than the one typed.
+      ["0x1", "0x1"],
+      ["1e999", "1e999"],
+    ] as const) {
+      expect(
+        await bodiesOf(["--index", "movies", "--ranking-score-threshold", text]),
+        `--ranking-score-threshold ${text}`,
+      ).toEqual([{ rankingScoreThreshold: sent }]);
+    }
+  });
+
+  it("refuses --output ndjson: the response is hits, not a list envelope", async () => {
+    server.use(meilisearchHandlers.search(hits));
+    const result = await runCommand(Search, ["--index", "movies", "--output", "ndjson"], { env });
+    expect(result.exitCode).toBe(2);
+    expect(result.stdout).toBe("");
+    expect(JSON.parse(result.stderr).error.code).toBe("cli_usage");
+  });
+
+  it("passes the API error of a missing index through and exits 4", async () => {
+    server.use(
+      meilisearchHandlers.apiError("/indexes/missing/search", 404, {
+        message: "Index `missing` not found.",
+        code: "index_not_found",
+        type: "invalid_request",
+        link: "https://docs.meilisearch.com/errors#index_not_found",
+      }),
+    );
+    const result = await runCommand(Search, ["--index", "missing", "--output", "json"], { env });
+    expect(result).toMatchSnapshot();
+  });
+});
diff --git a/test/contract/__snapshots__/commands.test.ts.snap b/test/contract/__snapshots__/commands.test.ts.snap
index b205df4..9a69ec5 100644
--- a/test/contract/__snapshots__/commands.test.ts.snap
+++ b/test/contract/__snapshots__/commands.test.ts.snap
@@ -708,6 +708,227 @@ The file holds API keys and decides where every later command goes, for every pr
       ],
       "id": "schema",
     },
+    {
+      "api": "meilisearch",
+      "description": "Search the documents of an index, with --q and the search parameters of the route.",
+      "examples": [
+        "meili search --index movies --q nosferatu",
+        "meili search --index movies --q nosferatu --output json",
+        "meili search --index movies --q horror --filter 'rating > 4' --sort rating:desc --limit 5",
+        "meili search --index movies --facets genres --attributes-to-retrieve id,title",
+        "meili search --index movies --data '{"q":"nosferatu","hybrid":{"embedder":"default","semanticRatio":0.5}}'",
+      ],
+      "flags": [
+        {
+          "description": "Project URL, the Meilisearch instance to talk to.",
+          "helpValue": "<url>",
+          "name": "project-url",
+          "type": "option",
+        },
+        {
+          "description": "Project API key. \`-\` reads it from stdin.",
+          "helpValue": "<key>",
+          "name": "project-api-key",
+          "type": "option",
+        },
+        {
+          "description": "A project of the config file. Cannot be combined with --project-url. Required when the file has several projects.",
+          "helpValue": "<name>",
+          "name": "project",
+          "type": "option",
+        },
+        {
+          "description": "The index to search in.",
+          "helpValue": "<index>",
+          "name": "index",
+          "required": true,
+          "type": "option",
+        },
+        {
+          "description": "The search terms. Omit it for a placeholder search, which returns everything.",
+          "helpValue": "<query>",
+          "name": "q",
+          "type": "option",
+        },
+        {
+          "description": "Number of documents to skip. Ignored when --page or --hits-per-page is given.",
+          "helpValue": "<n>",
+          "name": "offset",
+          "type": "option",
+        },
+        {
+          "description": "Maximum number of documents to return, 20 by default. Ignored when --page or --hits-per-page is given.",
+          "helpValue": "<n>",
+          "name": "limit",
+          "type": "option",
+        },
+        {
+          "description": "Page to return, 1-indexed. The response then carries totalHits and totalPages instead of estimatedTotalHits.",
+          "helpValue": "<n>",
+          "name": "page",
+          "type": "option",
+        },
+        {
+          "description": "Number of documents per page, used with --page. 0 counts without returning any.",
+          "helpValue": "<n>",
+          "name": "hits-per-page",
+          "type": "option",
+        },
+        {
+          "description": "Document attributes to return, comma-separated. Defaults to the displayed ones.",
+          "helpValue": "<fields>",
+          "name": "attributes-to-retrieve",
+          "type": "option",
+        },
+        {
+          "description": "Attributes to crop in _formatted, comma-separated, each optionally as \`attribute:length\`.",
+          "helpValue": "<fields>",
+          "name": "attributes-to-crop",
+          "type": "option",
+        },
+        {
+          "description": "Maximum number of words in a cropped value, 10 by default.",
+          "helpValue": "<n>",
+          "name": "crop-length",
+          "type": "option",
+        },
+        {
+          "description": "String marking where a cropped value was cut, \`…\` by default.",
+          "helpValue": "<string>",
+          "name": "crop-marker",
+          "type": "option",
+        },
+        {
+          "description": "Attributes to highlight in _formatted, comma-separated.",
+          "helpValue": "<fields>",
+          "name": "attributes-to-highlight",
+          "type": "option",
+        },
+        {
+          "description": "String inserted before each highlighted term, \`<em>\` by default.",
+          "helpValue": "<string>",
+          "name": "highlight-pre-tag",
+          "type": "option",
+        },
+        {
+          "description": "String inserted after each highlighted term, \`</em>\` by default.",
+          "helpValue": "<string>",
+          "name": "highlight-post-tag",
+          "type": "option",
+        },
+        {
+          "description": "Add _matchesPosition to each hit, the byte offset of every matched term.",
+          "name": "show-matches-position",
+          "type": "boolean",
+        },
+        {
+          "description": "Filter expression, for example \`rating > 4\`. Its array-of-arrays form goes through --data.",
+          "helpValue": "<filter>",
+          "name": "filter",
+          "type": "option",
+        },
+        {
+          "description": "Attributes to sort on, comma-separated, each as \`attribute:asc\` or \`:desc\`.",
+          "helpValue": "<sort>",
+          "name": "sort",
+          "type": "option",
+        },
+        {
+          "description": "Return one document per value of this attribute, which must be filterable.",
+          "helpValue": "<attribute>",
+          "name": "distinct",
+          "type": "option",
+        },
+        {
+          "description": "Attributes to count in facetDistribution, comma-separated. Patterns such as \`*\` are accepted.",
+          "helpValue": "<facets>",
+          "name": "facets",
+          "type": "option",
+        },
+        {
+          "description": "How query terms are relaxed when there are not enough results.",
+          "name": "matching-strategy",
+          "options": [
+            "last",
+            "all",
+            "frequency",
+          ],
+          "type": "option",
+        },
+        {
+          "description": "Restrict the search to these searchable attributes, comma-separated.",
+          "helpValue": "<fields>",
+          "name": "attributes-to-search-on",
+          "type": "option",
+        },
+        {
+          "description": "Drop the hits whose ranking score is below this value, between 0.0 and 1.0.",
+          "helpValue": "<number>",
+          "name": "ranking-score-threshold",
+          "type": "option",
+        },
+        {
+          "description": "ISO-639 locales of the query, comma-separated, instead of auto-detection.",
+          "helpValue": "<locales>",
+          "name": "locales",
+          "type": "option",
+        },
+        {
+          "description": "Hybrid search as JSON, for example \`{"embedder":"default","semanticRatio":0.5}\`.",
+          "helpValue": "<json>",
+          "name": "hybrid",
+          "type": "option",
+        },
+        {
+          "description": "Query vector for a user-provided embedder, comma-separated numbers.",
+          "helpValue": "<numbers>",
+          "name": "vector",
+          "type": "option",
+        },
+        {
+          "description": "Add _vectors to each hit, the embeddings of the document.",
+          "name": "retrieve-vectors",
+          "type": "boolean",
+        },
+        {
+          "description": "Add _rankingScore to each hit, between 0.0 and 1.0.",
+          "name": "show-ranking-score",
+          "type": "boolean",
+        },
+        {
+          "description": "Add _rankingScoreDetails to each hit, the score of every ranking rule.",
+          "name": "show-ranking-score-details",
+          "type": "boolean",
+        },
+        {
+          "description": "Add performanceDetails to the response, a timing breakdown of the query.",
+          "name": "show-performance-details",
+          "type": "boolean",
+        },
+        {
+          "description": "Read the body from a file. \`-\` reads from stdin.",
+          "exclusive": [
+            "data",
+          ],
+          "helpValue": "<path>",
+          "name": "file",
+          "type": "option",
+        },
+        {
+          "description": "Inline JSON body.",
+          "helpValue": "<json>",
+          "name": "data",
+          "type": "option",
+        },
+      ],
+      "id": "search",
+      "keyActions": [
+        "search",
+      ],
+      "routes": [
+        "POST /indexes/{index_uid}/search",
+      ],
+    },
     {
       "api": "meilisearch",
       "async": true,
@@ -973,6 +1194,13 @@ The file holds API keys and decides where every later command goes, for every pr
         "description": "Print one line per command, or one command in full with --command, from the oclif manifest.",
         "id": "schema",
       },
+      {
+        "description": "Search the documents of an index, with --q and the search parameters of the route.",
+        "id": "search",
+        "routes": [
+          "POST /indexes/{index_uid}/search",
+        ],
+      },
       {
         "description": "Add documents to an index, merging their fields into the ones that already have the same primary key.",
         "id": "update documents",
diff --git a/test/helpers/meilisearch-handlers.ts b/test/helpers/meilisearch-handlers.ts
index 4125bbc..6c62e94 100644
--- a/test/helpers/meilisearch-handlers.ts
+++ b/test/helpers/meilisearch-handlers.ts
@@ -99,6 +99,24 @@ export const meilisearchHandlers = {
       HttpResponse.json(task),
     ),
 
+  /**
+   * `POST /indexes/{uid}/search` answering the envelope of the API at its
+   * default pagination, with the `q` it was sent. `search` has no `--all`, and
+   * the tests that pass `--limit` assert the request body, not the response.
+   */
+  search: (hits: RecordAny[]) =>
+    http.post(`${MEILISEARCH_URL}/indexes/:uid/search`, async ({ request }) => {
+      const { q } = (await request.json()) as { q?: string };
+      return HttpResponse.json({
+        hits,
+        query: q ?? "",
+        processingTimeMs: 1,
+        limit: 20,
+        offset: 0,
+        estimatedTotalHits: hits.length,
+      });
+    }),
+
   /** Any method on `path` answering a Meilisearch error body with `status`. */
   apiError: (path: string, status: number, body: MeilisearchErrorResponse) =>
     http.all(`${MEILISEARCH_URL}${path}`, () => HttpResponse.json(body, { status })),
diff --git a/test/integration/meilisearch/search.test.ts b/test/integration/meilisearch/search.test.ts
new file mode 100644
index 0000000..45e08dc
--- /dev/null
+++ b/test/integration/meilisearch/search.test.ts
@@ -0,0 +1,132 @@
+import { randomUUID } from "node:crypto";
+import { Meilisearch, type SearchResponse } from "meilisearch";
+import { afterAll, beforeAll, describe, expect, it } from "vitest";
+import { runBinary } from "../../helpers/run-binary.js";
+import { MASTER_KEY, MEILISEARCH_URL, meilisearchAvailable } from "./setup.js";
+
+const meilisearchUp = await meilisearchAvailable();
+
+const env = { MEILI_PROJECT_URL: MEILISEARCH_URL, MEILI_PROJECT_API_KEY: MASTER_KEY };
+const uid = `meili-cli-test-search-${randomUUID().slice(0, 8)}`;
+const documents = [
+  { id: 1, title: "Nosferatu", genre: "horror", rating: 4.5 },
+  { id: 2, title: "Nosferatu the Vampyre", genre: "horror", rating: 4 },
+  { id: 3, title: "Solaris", genre: "scifi", rating: 4.8 },
+];
+
+describe.skipIf(!meilisearchUp)("meili search against a real Meilisearch", () => {
+  const client = new Meilisearch({ host: MEILISEARCH_URL, apiKey: MASTER_KEY });
+
+  beforeAll(async () => {
+    await client.createIndex(uid, { primaryKey: "id" }).waitTask();
+    await client
+      .index(uid)
+      .updateSettings({
+        filterableAttributes: ["genre", "rating"],
+        sortableAttributes: ["rating"],
+      })
+      .waitTask();
+    await client.index(uid).addDocuments(documents).waitTask();
+  });
+
+  afterAll(async () => {
+    await client.deleteIndex(uid).waitTask();
+  });
+
+  it("returns the hits of --q with the API envelope", async () => {
+    const result = await runBinary(
+      ["search", "--index", uid, "--q", "nosferatu", "--output", "json"],
+      env,
+    );
+    expect(result.exitCode).toBe(0);
+    expect(result.stderr).toBe("");
+    const body = JSON.parse(result.stdout) as SearchResponse;
+    expect(body.query).toBe("nosferatu");
+    expect(body.hits.map((hit) => hit.id)).toEqual([1, 2]);
+    expect(body.estimatedTotalHits).toBe(2);
+  });
+
+  it("returns every document for a placeholder search", async () => {
+    const result = await runBinary(["search", "--index", uid, "--output", "json"], env);
+    expect(result.exitCode).toBe(0);
+    expect((JSON.parse(result.stdout) as SearchResponse).hits).toHaveLength(3);
+  });
+
+  it("filters, sorts, projects and facets", async () => {
+    const result = await runBinary(
+      [
+        "search",
+        "--index",
+        uid,
+        "--filter",
+        "rating > 4",
+        "--sort",
+        "rating:desc",
+        "--attributes-to-retrieve",
+        "id",
+        "--facets",
+        "genre",
+        "--output",
+        "json",
+      ],
+      env,
+    );
+    expect(result.exitCode).toBe(0);
+    const body = JSON.parse(result.stdout) as SearchResponse;
+    expect(body.hits).toEqual([{ id: 3 }, { id: 1 }]);
+    expect(body.facetDistribution).toEqual({ genre: { horror: 1, scifi: 1 } });
+  });
+
+  it("switches the envelope to totalHits and totalPages with --page", async () => {
+    const result = await runBinary(
+      ["search", "--index", uid, "--page", "2", "--hits-per-page", "2", "--output", "json"],
+      env,
+    );
+    expect(result.exitCode).toBe(0);
+    const body = JSON.parse(result.stdout) as SearchResponse;
+    expect(body.hits).toHaveLength(1);
+    expect(body.totalHits).toBe(3);
+    expect(body.totalPages).toBe(2);
+  });
+
+  it("sends the body of --data, which a flag of the same field overrides", async () => {
+    const result = await runBinary(
+      [
+        "search",
+        "--index",
+        uid,
+        "--data",
+        '{"q":"solaris","limit":1}',
+        "--q",
+        "nosferatu",
+        "--output",
+        "json",
+      ],
+      env,
+    );
+    expect(result.exitCode).toBe(0);
+    const body = JSON.parse(result.stdout) as SearchResponse;
+    expect(body.query).toBe("nosferatu");
+    expect(body.hits).toHaveLength(1);
+  });
+
+  it("lets the API refuse a flag value that does not convert, with its own code", async () => {
+    const result = await runBinary(
+      ["search", "--index", uid, "--ranking-score-threshold", "abc", "--output", "json"],
+      env,
+    );
+    expect(result.exitCode).toBe(1);
+    expect(result.stdout).toBe("");
+    expect(JSON.parse(result.stderr).error.code).toBe("invalid_search_ranking_score_threshold");
+  });
+
+  it("exits 4 with the API error when the index does not exist", async () => {
+    const result = await runBinary(
+      ["search", "--index", `${uid}-missing`, "--q", "nosferatu", "--output", "json"],
+      env,
+    );
+    expect(result.exitCode).toBe(4);
+    expect(result.stdout).toBe("");
+    expect(JSON.parse(result.stderr).error.code).toBe("index_not_found");
+  });
+});