--- title: "Configuration" description: "Nuxt configuration reference for Nuxt AI Ready." canonical_url: "https://nuxtseo.com/docs/ai-ready/api/config" last_updated: "2026-09-25T20:11:21.282Z" --- Set module options under `aiReady` in your Nuxt configuration. Defaults below describe the current module. ```ts [nuxt.config.ts] export default defineNuxtConfig({ aiReady: { // options } }) ``` ## `enabled: boolean`{lang="ts"} - Default: `true`{lang="ts"} ```ts aiReady: { enabled: process.env.NODE_ENV === 'production' } ``` ## `debug: boolean`{lang="ts"} - Default: `false`{lang="ts"} Enable debug logging for module operations. ## `contentNegotiation: boolean`{lang="ts"} - Default: Automatic per route Set `false`{lang="ts"} to keep original page URLs as HTML, even when requests prefer Markdown through `Accept`{lang="http"} or an AI crawler `User-Agent`{lang="http"}. Explicit `.md` URLs, alternate links, `llms.txt`, MCP, and page indexing remain available. In automatic mode, the module reads the effective route rule for each request. It disables negotiation for ISR and for Nitro response caches whose `varies`{lang="ts"} list is missing `accept`{lang="http"}, `sec-fetch-dest`{lang="http"}, or `user-agent`{lang="http"}. This includes `swr`{lang="ts"} rules after Nitro normalizes them. `headersOnly: true`{lang="ts"} remains safe because it does not cache the response body. Set `false`{lang="ts"} for URL-only CDN or reverse proxy caching configured outside Nuxt. Set `true`{lang="ts"} to force negotiation and take responsibility for the cache key. ```ts export default defineNuxtConfig({ aiReady: { contentNegotiation: false } }) ``` ## `apiCatalog: false | ApiCatalogConfig`{lang="ts"} - Default: Disabled until you configure entries or an integration generates them Publish an [RFC 9727 API Catalog](/docs/ai-ready/guides/api-catalog) at `/.well-known/api-catalog`{lang="bash"}. Relative URLs resolve against `site.url`{lang="ts"} and `app.baseURL`{lang="ts"} once during module setup. Non-root apps redirect the origin-root request to the base-aware route. ```ts export default defineNuxtConfig({ site: { url: 'https://example.com', }, aiReady: { apiCatalog: { entries: [{ anchor: '/api', serviceDesc: { href: '/openapi.json', type: 'application/vnd.oai.openapi+json;version=3.1', }, serviceDoc: { href: '/docs/api', type: 'text/html' }, status: { href: '/api/health', type: 'application/json' }, }], }, }, }) ``` Each entry needs at least one relation target. A relation accepts one target or an array. Supported typed relations are `item`{lang="ts"}, `serviceDesc`{lang="ts"}, `serviceDoc`{lang="ts"}, `serviceMeta`{lang="ts"}, `status`{lang="ts"}, and `apiCatalog`{lang="ts"}. Use `relations`{lang="ts"} for other RFC 8288 relation names. Set `apiCatalog: false`{lang="ts"} to suppress configured and integration-generated entries. ## `mdreamOptions: MdreamOptions`{lang="ts"} - Default: `{ minimal: true, clean: true }`{lang="ts"} Configure [mdream](https://github.com/harlan-zw/mdream) HTML-to-markdown conversion. ```ts export default defineNuxtConfig({ aiReady: { mdreamOptions: { minimal: true } } }) ``` ## `markdownCacheHeaders: { maxAge?: number, swr?: boolean }`{lang="ts"} - Default: `{ maxAge: 3600, swr: true }`{lang="ts"} Cache settings for runtime markdown endpoints. | Option | Type | Default | Description | | -------- | --------- | ------- | ------------------------- | | `maxAge` | `number` | `3600` | Cache duration in seconds | | `swr` | `boolean` | `true` | Stale-while-revalidate | ## `agentSkills: false | AgentSkillsConfig`{lang="ts"} - Default: `{}`{lang="ts"} Publish an [Agent Skills Discovery v0.2.0 index](/docs/ai-ready/guides/agent-skills). The module finds `skills//SKILL.md`{lang="text"} files in your project and its layers. It serves each file at `/.well-known/agent-skills//SKILL.md`{lang="text"} and `/skills//SKILL.md`{lang="text"}. For one skill, or the skill selected by `root`, it also serves `/SKILL.md`{lang="text"}. The module lists discovered skills in llms.txt. With no skills, it publishes nothing. Use `dir` to change the folder or disable discovery. Use `skills` to add local or external entries. Set `llmsTxt: false` to leave llms.txt alone, or `agentSkills: false` to disable publication. The `ai-ready:agent-skills` hook sees every entry before publication. ```ts aiReady: { agentSkills: { skills: [{ source: 'local', name: 'seo-audit', description: 'Audit a site for critical SEO issues.', file: './skills/seo-audit/SKILL.md' }] } } ``` ## `llmsTxt: LlmsTxtConfig`{lang="ts"} - Default: Auto-generated Configure llms.txt generation. ```ts export default defineNuxtConfig({ aiReady: { llmsTxt: { markdownLinks: true, sections: [ { title: 'API Reference', links: [ { title: 'REST API', href: '/docs/api', description: 'API docs' } ] }, { title: 'Debug Endpoints', optional: true, links: [ { title: 'Debug Route', href: '/__ai-ready-debug', description: 'Internal debugging' } ] } ], notes: 'Built with Nuxt AI Ready' } } }) ``` **LlmsTxtSection:** | Property | Type | Description | | ------------- | --------------------------------- | --------------------------------------------------------------------------------------------------------------- | | `title` | `string` | Section title | | `description` | `string \| string[]` | Section context rendered in the heading-free preamble, or in each optional link note | | `links` | `{ title, href, description? }[]` | Links in section | | `optional` | `boolean` | Put the section's links in the single `## Optional` file list, which LLMs may skip with shorter context windows | **LlmsTxtConfig:** | Property | Type | Description | | --------------- | -------------------- | --------------------------------------------------------------------------------------------------------------- | | `markdownLinks` | `boolean` | Link automatically generated page and locale entries to available Markdown representations. Defaults to `false` | | `sections` | `LlmsTxtSection[]` | Custom sections | | `notes` | `string \| string[]` | Additional heading-free context rendered before file-list sections | ## `sitemapMd: boolean`{lang="ts"} - Default: `true`{lang="ts"} Generate a Markdown sitemap at `/sitemap.md`{lang="bash"} from indexed pages. The sitemap groups pages by top-level path segment under `##`{lang="ts"} headings. Each link points at the page’s `.md` representation. When a page records a last update, it becomes the link's title text. The same flag appends a `## Sitemap`{lang="ts"} section to module-generated Markdown, linking back to `/sitemap.md`{lang="bash"}. ```ts export default defineNuxtConfig({ aiReady: { sitemapMd: false } }) ``` ## `describedby: boolean`{lang="ts"} - Default: `true`{lang="ts"} Advertise llms.txt with `rel="describedby"`{lang="ts"} on HTML and Markdown responses. HTML pages get a ``{lang="html"} tag, and both variants get a `Link`{lang="http"} header entry. ```ts export default defineNuxtConfig({ aiReady: { describedby: false } }) ``` ## `contentSignal: false | { aiTrain?: boolean, contentUsage?: boolean, search?: boolean, aiInput?: boolean }`{lang="ts"} - Default: `false`{lang="ts"} Content Signal directives for robots.txt. See [Content Signals guide](/docs/ai-ready/guides/content-signals). ```ts export default defineNuxtConfig({ aiReady: { contentSignal: { aiTrain: false, // Block training contentUsage: false, // Omit Content-Usage from robots.txt search: true, // Allow search indexing aiInput: true, // Allow RAG/grounding } } }) ``` ## `tools: SiteToolsConfig`{lang="ts"} Configure the three built-in tools once, then attach each one to MCP, WebMCP or both. | Tool | Shared options | `mcp` options | `webmcp` options | | ----------------- | -------------- | ------------- | ---------------------------------------- | | `listPages` | `defaultLimit` | `enabled` | `enabled`, `maxOutputChars`, `exposedTo` | | `searchPages` | `defaultLimit` | `enabled` | `enabled`, `maxOutputChars`, `exposedTo` | | `getPageMarkdown` | None | `enabled` | `enabled`, `maxOutputChars`, `exposedTo` | `defaultLimit` defaults to 20 for `listPages` and 10 for `searchPages`. Values are clamped to 50. WebMCP output defaults to 1,500 characters per tool. ```ts export default defineNuxtConfig({ aiReady: { tools: { listPages: { defaultLimit: 10, webmcp: { enabled: false }, }, searchPages: { defaultLimit: 5, webmcp: { maxOutputChars: 3000 }, }, }, }, }) ``` The module enables omitted attachment objects when their transport is active. Set `mcp.enabled`{lang="ts"} or `webmcp.enabled`{lang="ts"} to `false` on one tool to leave it off that transport. ## `mcp: { tools?: boolean, resources?: boolean }`{lang="ts"} - Default: `{ tools: true, resources: true }`{lang="ts"} Control MCP features when `@nuxtjs/mcp-toolkit` installed. ```ts export default defineNuxtConfig({ aiReady: { mcp: { tools: true, // list_pages, search_pages, get_page_markdown resources: true, // pages resource } } }) ``` See [MCP guide](/docs/ai-ready/guides/mcp) for tool/resource details. ## `mcpServerCard: false | McpServerCardConfig`{lang="ts"} - Default: `{ cacheMaxAge: 3600 }`{lang="ts"} Publish experimental SEP-2127 discovery metadata at `/server-card`{lang="bash"} when MCP Toolkit runs on the server. For the default Toolkit route, the card is available at `/mcp/server-card`{lang="bash"}. Set this option to `false` to disable the route. | Option | Type | Default | Description | | ------------- | -------- | ------------------------------------ | --------------------------------------------------- | | `name` | `string` | Derived from site host and MCP route | Reverse-DNS/server identity | | `title` | `string` | Existing Toolkit or site name | Human-readable server title | | `description` | `string` | Toolkit or site description | Server description, up to 100 characters | | `websiteUrl` | `string` | None | Absolute HTTP or HTTPS project or documentation URL | | `cacheMaxAge` | `number` | `3600` | Browser and shared cache lifetime in seconds | Toolkit icons are copied into the card. The card route is omitted when Toolkit is absent, disabled or unavailable in a static deployment. ## `webmcp: boolean | WebMcpConfig`{lang="ts"} - Default: `false`{lang="ts"} Register [WebMCP](/docs/ai-ready/guides/webmcp) tools for browser agents through `document.modelContext`{lang="ts"}. | Option | Type | Default | Description | | ----------- | ---------- | ------- | -------------------------------------------------------------------------------------- | | `tools` | `boolean` | `true` | Attach tools enabled under `aiReady.tools` | | `exposedTo` | `string[]` | None | Default trusted origins for built-in and composable tools. Same-origin only by default | ```ts export default defineNuxtConfig({ aiReady: { tools: { listPages: { webmcp: { enabled: false }, }, searchPages: { defaultLimit: 20, webmcp: { maxOutputChars: 3000 }, }, }, webmcp: { exposedTo: ['https://agent.example.com'], }, }, }) ``` Nuxt auto-imports `useWebMcpTool()`{lang="ts"} and `useWebMcpSupported()`{lang="ts"} when you enable WebMCP. Set `webmcp.tools`{lang="ts"} to `false` to register your own tools without the built-in definitions or `/__ai-ready/pages`{lang="bash"} endpoint. ## `llmsTxtCacheSeconds: number`{lang="ts"} - Default: `600`{lang="ts"} (10 minutes) Cache duration for llms.txt route handlers. Uses stale-while-revalidate. ```ts export default defineNuxtConfig({ aiReady: { llmsTxtCacheSeconds: 3600 // 1 hour } }) ``` ## `database: false | { type?: string, filename?: string, bindingName?: string, url?: string, authToken?: string }`{lang="ts"} - Default: Disabled until a requested feature needs stored pages. Configure the database for page storage. The module auto-detects the best driver based on the deployment platform. | Option | Type | Default | Description | | ------------- | ------------------------------------------------------------------------- | --------------------------- | --------------------------------------------- | | `type` | `'none' \| 'sqlite' \| 'bun' \| 'd1' \| 'libsql' \| 'neon' \| 'postgres'` | `'sqlite'` | Database type. `'none'` disables the database | | `filename` | `string` | `'.data/ai-ready/pages.db'` | [SQLite](https://sqlite.org) file path | | `bindingName` | `string` | `'DB'` | D1 binding name | | `url` | `string` | - | LibSQL, Neon, or PostgreSQL URL | | `authToken` | `string` | - | LibSQL/Turso auth token | ::code-group ```ts [Default (SQLite)] export default defineNuxtConfig({ aiReady: { database: { filename: '.data/ai-ready/pages.db' } } }) ``` ```ts [Cloudflare D1] export default defineNuxtConfig({ aiReady: { database: { type: 'd1', bindingName: 'DB' } } }) ``` ```ts [Turso/LibSQL] export default defineNuxtConfig({ aiReady: { database: { type: 'libsql', url: process.env.TURSO_URL, authToken: process.env.TURSO_AUTH_TOKEN } } }) ``` ```ts [PostgreSQL] export default defineNuxtConfig({ aiReady: { database: { type: 'postgres', url: process.env.DATABASE_URL } } }) ``` ```ts [No database] export default defineNuxtConfig({ aiReady: { database: false } }) ``` :: **Driver auto-detection:** | Platform | Driver | Extra dependency | | -------------------------------------------------- | ---------------- | ---------------- | | [Node.js](https://nodejs.org) 22.13+ | `node:sqlite` | None | | Earlier [Node.js](https://nodejs.org) | `better-sqlite3` | `better-sqlite3` | | [Bun](https://bun.sh) | `bun:sqlite` | None | | [Cloudflare](https://cloudflare.com) | D1 | None | | [Vercel](https://vercel.com) (with `POSTGRES_URL`) | Neon Postgres | None | | Node.js with `database.type: 'postgres'` | PostgreSQL | `postgres` | On Node.js before 22.13, the default `sqlite` type requires the optional [`better-sqlite3`](https://github.com/WiseLibs/better-sqlite3) peer dependency: `pnpm add better-sqlite3`{lang="bash"}. For [PostgreSQL](https://postgresql.org), install the optional [`postgres`](https://github.com/porsager/postgres) peer dependency: `pnpm add postgres`{lang="bash"}. The driver reads `database.url`{lang="ts"}, `POSTGRES_URL`, then `DATABASE_URL`. Prerendering creates a compressed dump. Runtime storage restores it when empty and checks build changes on later requests. ### No Database The database stays off when no requested feature needs stored pages. Set `database: false`{lang="ts"} (or `database: { type: 'none' }`{lang="ts"}) to force it off. No driver enters the server bundle, so the runtime needs no `better-sqlite3`. On Node.js before 22.13, build time indexing needs `better-sqlite3`. Runtime indexing, MCP page tools or resources, and WebMCP page tools enable storage automatically. An explicit database config also enables storage. Build-time generation still works. You keep: - `llms.txt` and `llms-full.txt`, written during prerender. - `.md` twins for every prerendered page. - The `Link` headers, configured content signals, API Catalog and Agent Skills. Runtime Markdown conversion remains available without a database when the deployment retains a server. A fully static deployment can only serve its generated files. The following features need stored pages: - `list_pages`, `search_pages` and `get_page_markdown`, for both MCP and WebMCP. - The MCP pages resource. - Runtime `llms-full.txt` needs stored pages to include page bodies. Prerender the route for static output. A disabled database conflicts with `runtimeSync` and `cron`. If you set any of them, the build fails with a message naming the option. **Table naming:** The module uses these reserved table names: - `ai_ready_pages` : Main pages table - `ai_ready_pages_fts` : FTS5 full-text search index - `_ai_ready_info` : Schema version tracking The prefixes separate these tables from unrelated application tables. Do not reuse the same table names for other data. ## `cron: boolean`{lang="ts"} - Default: `false`{lang="ts"} Enable the scheduled task that runs every 5 minutes. It automatically enables `runtimeSync` for background indexing. ```ts export default defineNuxtConfig({ aiReady: { cron: true } }) ``` ## `runtimeSyncSecret: string`{lang="ts"} - Default: Generated when you enable `runtimeSync` or `cron` Secret token for runtime control endpoints. Send it in the `Authorization: Bearer `{lang="http"} header. Setup uses explicit config, then `NUXT_AI_READY_RUNTIME_SYNC_SECRET`, then a generated value when you enable runtime sync. Use a stable configured secret for external schedulers and the CLI. Setup writes the chosen secret into the local CLI cache. Status requests also authenticate; prune dry runs are the exception. ```ts export default defineNuxtConfig({ aiReady: { runtimeSyncSecret: process.env.AI_READY_SECRET } }) ``` ## `runtimeSync: boolean | { ttl?: number, batchSize?: number, pruneTtl?: number }`{lang="ts"} - Default: `false`{lang="ts"} Enable runtime controls and sitemap seeding for content that changes between deployments. Set `true` for defaults, or use an object to change them. Poll and cron process pending pages. TTL expiry alone does not mark an already indexed page pending. When a published page changes between builds, call the authenticated reindex endpoint or a manual indexing utility. | Option | Type | Default | Description | | ----------- | -------- | ------- | --------------------------------------------------------------------------- | | `ttl` | `number` | `3600` | Freshness window for manual indexing and sitemap refresh interval (seconds) | | `batchSize` | `number` | `50` | Pages per batch (max: 50) | | `pruneTtl` | `number` | `0` | Prune routes not seen in the sitemap for this many seconds (0 = disabled) | ```ts export default defineNuxtConfig({ aiReady: { runtimeSync: { ttl: 3600, batchSize: 20, pruneTtl: 604800 // prune after 7 days }, runtimeSyncSecret: process.env.AI_READY_SECRET, cron: true // enable scheduled indexing } }) ``` **When enabled:** - `GET /__ai-ready/status` : Check indexing progress - `POST /__ai-ready/poll` : Trigger batch indexing (requires `Authorization: Bearer `{lang="html"} header) - `POST /__ai-ready/prune` : Prune stale routes (requires `Authorization: Bearer `{lang="html"} header) - `POST /__ai-ready/reindex?route=/about`{lang="bash"} - Reindex a single route (requires `Authorization: Bearer `{lang="html"} header) See [Runtime Sync guide](/docs/ai-ready/guides/runtime-indexing) for details. ## `autoI18n: boolean`{lang="ts"} - Default: `true`{lang="ts"} Auto-detect [`@nuxtjs/i18n`](https://i18n.nuxtjs.org/) (or `nuxt-i18n-micro`) at build time and integrate locale data into: - **`llms.txt`**: adds an `## Available Languages on Website` file-list section linking to every locale root - **`Link` headers**: emits `Link: <…>; rel="alternate"; hreflang="…"` per locale on `.md` and HTML responses (RFC 8288) - **Frontmatter**: runtime HTML-to-Markdown responses include a `locale` field - **Database**: each indexed page persists its `locale` for filtered queries - **FTS5 tokenizer**: auto-switches to `trigram` for CJK locales (zh / ja / ko) Disable to skip the integration even with an i18n module present: ```ts export default defineNuxtConfig({ aiReady: { autoI18n: false } }) ``` See [i18n guide](/docs/ai-ready/guides/i18n) for details. ## `contentSource: boolean`{lang="ts"} - Default: `true`{lang="ts"} Read Markdown directly from page collections in `@nuxt/content` v3 or `@harlan-zw/comark-content` when available. Set `false` to convert rendered HTML instead. Use that when a page’s meaning depends on its rendered components. See [Markdown Conversion](/docs/ai-ready/guides/markdown) for source lookup behavior. ## `prerender: { concurrency?: number }`{lang="ts"} - Default concurrency: `10`{lang="ts"} Control how many pages the module processes concurrently during prerendering. Set `1` to process one page at a time. ```ts [nuxt.config.ts] export default defineNuxtConfig({ aiReady: { prerender: { concurrency: 5 }, }, }) ``` ## `debugCron: boolean`{lang="ts"} - Default: `false`{lang="ts"} Record cron runs in the database for diagnostics. Enable it when investigating scheduled indexing. This option does not enable cron itself. ## Sitemap See the full [sitemap](/sitemap.md) for all pages.