> ## Documentation Index > Fetch the complete documentation index at: https://docs.linkup.so/llms.txt > Use this file to discover all available pages before exploring further. # Migrating from Exa > Guide for switching from Exa Search, Contents, and Answer to Linkup Search and Fetch ## Overview Switch from Exa to Linkup's [**Search**](/pages/documentation/endpoints/search/overview), [**Fetch**](/pages/documentation/endpoints/fetch/overview), and [**Research**](/pages/documentation/endpoints/research/overview) endpoints. | Exa | Linkup | Role | | ---------------- | ----------------------------------------------------------------------- | --------------------------------- | | `POST /search` | `POST /v1/search` | Web search for AI agents | | `POST /contents` | `POST /v1/fetch` | Clean content from a known URL | | `POST /answer` | `POST /v1/search` with `outputType` `"sourcedAnswer"` or `"structured"` | Cited answer to a question | | Exa Agent | `POST /v1/research` | Long-running, multi-step research | Most of the migration is a mechanical rename. The part that is not mechanical is the query itself. Exa and Linkup both take a query and return web results, but they expect the query to be written differently. Porting a call across without rewriting the query text does not give you a representative result. ## The core difference Both APIs put intelligence into the request, in different places. | | Exa | Linkup | | ------------------ | ---------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------- | | What the query is | A description of the page you want to find, often a short noun phrase. | An instruction to a search agent: what to find, and what to bring back. | | Where you steer it | Through parameters: `category`, `contents`, `systemPrompt`, `additionalQueries`, date filters. | Through the query text. The parameter surface is deliberately small. | | What you get back | The best-matching **page**. | The best-matching **fact**, with its source. | Anything you expressed to Exa as a parameter needs to be expressed to Linkup as words in `q`. A bare noun phrase gives the Linkup agent very little to plan a retrieval around, so rewriting the query is usually the difference between a mediocre result and a good one. **Migrating many call sites?** Give a coding agent a representative sample of your existing Exa calls plus [linkup-for-agents](https://github.com/LinkupPlatform/linkup-for-agents), our context pack of knowledge files, workflow recipes, and skills, and have it produce the translated versions. Install the skills into your project: ```bash theme={"system"} npx skills add LinkupPlatform/skills ``` See also [Linkup for agents](/pages/documentation/get-started/for-agents). ## Quick start ### Get your API key Create a Linkup account for free to get your API key. ### Install the SDK ```bash Python theme={"system"} pip install linkup-sdk ``` ```bash JavaScript theme={"system"} npm install linkup-sdk ``` ## The three settings that matter most Everything else is optional. These three account for nearly all of the difference in output quality. ### 1. `depth`: choose the retrieval strategy Exa's `type` is primarily a speed setting. Linkup's `depth` selects *how* the search is carried out, so map it deliberately rather than as a direct rename. `"flash"` and `"fast"` pass the query as-is to the index, with no LLM. `"standard"` interprets the query and can run parallel sub-searches. `"deep"` adds the ability to open a page it discovered, read it, and decide what to do next. If step two depends on what step one found, or a discovered page has to be read, use `"deep"`. | Exa `type` | Linkup `depth` | Use when | | ------------------------------------------- | -------------- | ------------------------------------------------------------------------------------- | | `"instant"` | `"flash"` | Lowest latency (\<200 ms). Ranked sources and snippets for chat, voice, autocomplete. | | `"fast"` | `"fast"` | One-shot retrieval where latency matters (\~1s). | | `"auto"` | `"standard"` | Everything you need can be searched for at once, in parallel (1–3s). | | `"deep-lite"`, `"deep"`, `"deep-reasoning"` | `"deep"` | Step two depends on step one, or a discovered page has to be opened and read (5–30s). | ### 2. `q`: write an instruction, not a description An Exa query describes a page. A Linkup query gives an instruction. Five moves that reliably improve results with `"standard"` and `"deep"`: 1. **Start with a verb.** Find, search, scrape, extract, compare. 2. **Name the fields you want back.** Linkup returns evidence rather than whole pages, so specify which evidence you need. 3. **Break breadth into explicit facets.** "Run separate searches for X, Y, and Z" is how you get parallel retrieval out of `"standard"`. This replaces Exa's `additionalQueries`. 4. **Repeat your constraints in every facet.** Without it, individual facets drift off-target. 5. **Ask for source URLs.** They make results verifiable at no extra cost. See [Search best practices](/pages/documentation/endpoints/search/best-practices) for more. ### 3. `outputType`: match it to what your code does next Exa returns pages and leaves the parsing to you. Linkup asks you to choose the response shape up front. | If your app… | Set `outputType` to | | ------------------------------------------- | -------------------------------------------------- | | Feeds results into your own LLM or reranker | `"searchResults"` | | Shows an answer to a person | `"sourcedAnswer"` | | Writes into a CRM, database, or spreadsheet | `"structured"` (requires `structuredOutputSchema`) | ## Migrate your search calls ### Replace your API calls **Exa** ```bash cURL theme={"system"} curl -X POST "https://api.exa.ai/search" \ -H "Content-Type: application/json" \ -H "x-api-key: YOUR_EXA_KEY" \ -d '{ "query": "latest AI news", "type": "auto", "numResults": 10, "contents": { "highlights": true } }' ``` ```python Python theme={"system"} from exa_py import Exa exa = Exa(api_key="YOUR_EXA_KEY") response = exa.search( "latest AI news", type="auto", num_results=10, contents={"highlights": True}, ) print(response) ``` ```javascript JavaScript theme={"system"} import Exa from "exa-js"; const exa = new Exa("YOUR_EXA_KEY"); const response = await exa.search("latest AI news", { type: "auto", numResults: 10, contents: { highlights: true } }); console.log(response); ``` **Linkup** ```bash cURL theme={"system"} curl -X POST "https://api.linkup.so/v1/search" \ -H "Authorization: Bearer YOUR_LINKUP_KEY" \ -H "Content-Type: application/json" \ -d '{ "q": "Find the latest AI news from the past week. Return the headline, publisher, and date for each story, with the source URL.", "depth": "standard", "outputType": "searchResults", "maxResults": 10 }' ``` ```python Python theme={"system"} from linkup import LinkupClient client = LinkupClient(api_key="YOUR_LINKUP_KEY") response = client.search( query="Find the latest AI news from the past week. Return the headline, publisher, and date for each story, with the source URL.", depth="standard", output_type="searchResults", max_results=10 ) print(response) ``` ```javascript JavaScript theme={"system"} import { LinkupClient } from 'linkup-sdk'; const client = new LinkupClient({ apiKey: 'YOUR_LINKUP_KEY' }); client.search({ query: 'Find the latest AI news from the past week. Return the headline, publisher, and date for each story, with the source URL.', depth: 'standard', outputType: 'searchResults', maxResults: 10 }).then(console.log); ``` ### Parameter mapping | Exa parameter | Linkup parameter | Notes | | ---------------------------- | ------------------------------------------------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | `query` | `q` | Required. **Rewrite it** as an instruction rather than copying it across. See the `q` section above | | `type` | `depth` | Required. Map `"instant"` → `"flash"`, `"fast"` → `"fast"`, `"auto"` → `"standard"`, `"deep-lite"` / `"deep"` / `"deep-reasoning"` → `"deep"` | | `numResults` | `maxResults` | Rename. `numResults` is not a Linkup parameter | | `category` | Express in `q` | `"company"` / `"people"`: "Find the official company page for…", or pass a LinkedIn URL directly. `"news"` / `"financial report"` / `"publication"`: "…from its most recent earnings release" | | `additionalQueries` | Express in `q` | "Run separate searches for: …". Works with `"standard"` and `"deep"` | | `includeDomains` | `includeDomains` | Array of domain strings, up to 100 entries (Exa allows 1,200). Reserve for domains you specifically need; a general preference such as "prefer official sources" works better in `q` | | `excludeDomains` | `excludeDomains` | Array of domain strings | | `startPublishedDate` | `fromDate` | ISO 8601 date (`YYYY-MM-DD`), not a full timestamp | | `endPublishedDate` | `toDate` | ISO 8601 date (`YYYY-MM-DD`), not a full timestamp | | `contents.highlights` | `outputType` `"searchResults"` | Linkup's default response shape. Each result includes a `content` snippet | | `contents.text` | **Fetch**, or `depth` `"deep"` | If you have the result URLs, call [**Fetch**](/pages/documentation/endpoints/fetch/overview) on them. To read pages the search discovers, use `"deep"` and say "scrape the page" in `q` | | `contents.summary` | `outputType` `"sourcedAnswer"` or `"structured"` | A summary schema maps to `structuredOutputSchema` | | `contents.subpages` | `depth` `"deep"` + instruction in `q` | For example: "if the page links to a pricing page, scrape that too" | | `contents.extras.imageLinks` | `includeImages` | Returns image results when supported | | `outputSchema` | `outputType` `"structured"` + `structuredOutputSchema` | JSON schema passed as a string; the root must be `type` `"object"`. See the [structured output guide](/pages/documentation/tutorials/structured-output-guide) | | `systemPrompt` | Express in `q` | "Prefer official sources" becomes a sentence in the query | | `userLocation` | Express in `q` | "…for the French market" | | `stream` | N/A | **Search** returns a single synchronous response | | `x-api-key` header | `Authorization: Bearer` header | — | ### Response format differences **Exa response structure** ```json theme={"system"} { "requestId": "b5947044c4b78efa9552a7c89b306d95", "results": [ { "title": "Page Title", "url": "https://example.com", "publishedDate": "2026-09-01T00:00:00.000Z", "highlights": ["Relevant excerpt..."] } ], "costDollars": { "total": 0.007 } } ``` **Linkup response structure (`"searchResults"`)** ```json theme={"system"} { "results": [ { "type": "text", "name": "Page Title", "url": "https://example.com", "content": "Relevant content..." } ] } ``` **Linkup response structure (`"sourcedAnswer"`)** ```json theme={"system"} { "answer": "Generated answer with citations...", "sources": [ { "name": "Page Title", "url": "https://example.com", "snippet": "Relevant excerpt..." } ] } ``` **Linkup response structure (`"structured"`)** ```json theme={"system"} { "customField1": "Extracted data following your schema...", "customField2": ["Array", "of", "values"] } ``` With `includeSources` set to `true`, the response becomes `{ "data": { ... }, "sources": [ ... ] }`. See the [structured output guide](/pages/documentation/tutorials/structured-output-guide). ### Worked example A typical Exa company search, translated. **Exa** ```json theme={"system"} { "query": "agtech companies in the US that have raised series A", "type": "auto", "category": "company", "numResults": 10, "contents": { "highlights": true } } ``` **Linkup** ```json theme={"system"} { "q": "Find US-based agtech companies that have raised a Series A round. Run separate searches for: US agtech startups that announced a Series A round; United States agriculture technology companies Series A funding round size and lead investor; American agtech Series A funding announcements in 2025 and 2026. For each company return the company name, headquarters, round size, announcement date, and lead investor, with the source URL for each finding.", "depth": "standard", "outputType": "searchResults", "maxResults": 10 } ``` What changed: * `category` `"company"` became words in the query. * `highlights` became `outputType` `"searchResults"`. * The noun phrase became an instruction with named fields. * One vague search became three parallel facets, with "US" and "Series A" repeated in all three so no facet drifts off-target. To write the results straight into a database, switch `outputType` to `"structured"` and pass a `structuredOutputSchema` describing one company per array item. ## Migrate your contents calls Exa **Contents** returns page content for known URLs. Linkup **Fetch** returns clean markdown for a single public URL (HTML or PDF), with optional JavaScript rendering and optional structured extraction. **Fetch** accepts one `url` per request. If you currently pass an array of URLs to Exa **Contents**, call Linkup **Fetch** once per URL (or in parallel). For batch workloads, submit your **Fetch** calls through [**Tasks**](/pages/documentation/endpoints/tasks/overview): up to 100 per submission, with the same parameters and pricing as direct `/fetch` calls. ### Replace your API calls **Exa** ```bash cURL theme={"system"} curl -X POST "https://api.exa.ai/contents" \ -H "Content-Type: application/json" \ -H "x-api-key: YOUR_EXA_KEY" \ -d '{ "urls": ["https://docs.linkup.so"], "text": true }' ``` ```python Python theme={"system"} from exa_py import Exa exa = Exa(api_key="YOUR_EXA_KEY") response = exa.get_contents( ["https://docs.linkup.so"], text=True ) print(response) ``` ```javascript JavaScript theme={"system"} import Exa from "exa-js"; const exa = new Exa("YOUR_EXA_KEY"); const response = await exa.getContents(["https://docs.linkup.so"], { text: true }); console.log(response); ``` **Linkup** ```bash cURL theme={"system"} curl -X POST "https://api.linkup.so/v1/fetch" \ -H "Authorization: Bearer YOUR_LINKUP_KEY" \ -H "Content-Type: application/json" \ -d '{ "url": "https://docs.linkup.so", "renderJs": true }' ``` ```python Python theme={"system"} from linkup import LinkupClient client = LinkupClient(api_key="YOUR_LINKUP_KEY") response = client.fetch( url="https://docs.linkup.so", render_js=True, ) print(response) ``` ```javascript JavaScript theme={"system"} import { LinkupClient } from 'linkup-sdk'; const client = new LinkupClient({ apiKey: 'YOUR_LINKUP_KEY' }); const response = await client.fetch({ url: 'https://docs.linkup.so', renderJs: true, }); console.log(response); ``` ### Parameter mapping | Exa parameter | Linkup parameter | Notes | | --------------------------- | ------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | | `urls` / `ids` | `url` | One URL per request. Batch through [**Tasks**](/pages/documentation/endpoints/tasks/overview) | | `text` | N/A | `markdown` is always returned | | `summary` (with `schema`) | `schema` + `instructions` | Returns typed JSON in `data` alongside the markdown. Fields not found on the page are omitted, not invented | | `extras.imageLinks` | `extractImages` | Returns a separate list of image URLs | | `maxAgeHours` / `livecrawl` | N/A | **Fetch** retrieves the page in real time | | `subpages` | N/A | **Fetch** reads one URL and does not follow links. Use **Search** with `depth` `"deep"`, or [**Extract**](/pages/documentation/endpoints/extract/overview) (closed beta) for listing pages | | N/A | `renderJs` | Render client-side JavaScript before extraction | | N/A | `mode` | `"pro"` for hard-to-retrieve pages | ### Response format differences **Exa response structure** ```json theme={"system"} { "requestId": "b5947044c4b78efa9552a7c89b306d95", "results": [ { "id": "https://docs.linkup.so", "url": "https://docs.linkup.so", "title": "Page Title", "text": "Extracted content..." } ], "statuses": [{ "id": "https://docs.linkup.so", "status": "success" }] } ``` **Linkup response structure** ```json theme={"system"} { "markdown": "# Page title\n\nExtracted content...", "images": [ { "alt": "Logo", "url": "https://example.com/logo.svg" } ] } ``` `markdown` is always returned. `images` is empty unless `extractImages` is `true`. `data` is present only when `schema` is set. ### Example: structured extraction from a known URL Replaces Exa **Contents** with a `summary` schema. ```bash cURL theme={"system"} curl -X POST "https://api.linkup.so/v1/fetch" \ -H "Authorization: Bearer YOUR_LINKUP_KEY" \ -H "Content-Type: application/json" \ -d '{ "url": "https://www.linkup.so/pricing", "renderJs": true, "schema": { "type": "object", "properties": { "plans": { "type": "array", "items": { "type": "object", "properties": { "name": { "type": "string" }, "price": { "type": "string", "description": "Public list price" } } } } } }, "instructions": "Express monetary values in USD." }' ``` ## Migrate your answer calls Exa **Answer** maps to Linkup **Search** with `outputType` `"sourcedAnswer"`. For a schema-shaped answer, use `outputType` `"structured"` with `structuredOutputSchema` in place of Exa's `outputSchema`. Fold `systemPrompt` and `userLocation` into `q`. **Exa** ```bash cURL theme={"system"} curl -X POST "https://api.exa.ai/answer" \ -H "Content-Type: application/json" \ -H "x-api-key: YOUR_EXA_KEY" \ -d '{ "query": "What is the latest valuation of SpaceX?", "systemPrompt": "Prefer official sources." }' ``` ```python Python theme={"system"} response = exa.answer( "What is the latest valuation of SpaceX?", system_prompt="Prefer official sources." ) ``` ```javascript JavaScript theme={"system"} const response = await exa.answer("What is the latest valuation of SpaceX?", { systemPrompt: "Prefer official sources." }); ``` **Linkup** ```bash cURL theme={"system"} curl -X POST "https://api.linkup.so/v1/search" \ -H "Authorization: Bearer YOUR_LINKUP_KEY" \ -H "Content-Type: application/json" \ -d '{ "q": "What is the latest valuation of SpaceX? Prefer official sources and recent reporting, and include the date of the valuation.", "depth": "standard", "outputType": "sourcedAnswer" }' ``` ```python Python theme={"system"} response = client.search( query="What is the latest valuation of SpaceX? Prefer official sources and recent reporting, and include the date of the valuation.", depth="standard", output_type="sourcedAnswer", ) print(response) ``` ```javascript JavaScript theme={"system"} const response = await client.search({ query: 'What is the latest valuation of SpaceX? Prefer official sources and recent reporting, and include the date of the valuation.', depth: 'standard', outputType: 'sourcedAnswer', }); console.log(response); ``` Exa returns `answer` plus `citations`. Linkup returns `answer` plus `sources`, each with `name`, `url`, and `snippet`. ## Migrate your agent calls For long-running research that Exa Agent handles (deep research, list building, enrichment), use Linkup [**Research**](/pages/documentation/endpoints/research/overview). It returns a sourced answer or structured output (`outputType` `"sourcedAnswer"` or `"structured"`), with `mode` and `reasoningDepth` to pin the type of investigation and its thoroughness. For list building and enrichment over many rows, also see [**Tasks**](/pages/documentation/endpoints/tasks/overview). ## Need help? * Check our [Quickstart Guide](/pages/documentation/get-started/quickstart) * Read [Search best practices](/pages/documentation/endpoints/search/best-practices) and [Fetch best practices](/pages/documentation/endpoints/fetch/best-practices) * Join our [Discord Community](https://discord.gg/9q9mCYJa86) * Explore the [Search](/pages/documentation/endpoints/search/reference) and [Fetch](/pages/documentation/endpoints/fetch/reference) API references Your Linkup account starts with \$20 of free credit when you sign up with a professional email address. You can monitor usage and add more credit in the [Billing](https://app.linkup.so/organization/billing) section.