workingproxysites.
[ new post ]
[ menu ]

directory

boards

[all]

information

Web-data service integration · Official publisher

Exa MCP

Use the default hosted connection for search and known-page fetching. Sign in or add an API key for Agent research. If you select tools explicitly, include every tool you need: the tools parameter replaces the defaults.

Server publisher: Exa. Guide maintained by Working Proxy Sites.

Documentation reviewed · Next review

What you can use it for

  • Discover source URLs and fetch a known page for closer reading.
  • Enable advanced search filters while preserving the tools the workflow needs.
  • Run authenticated research and resume an active run by its existing ID.

Documented capabilities: Web search · Scraping & extraction.

Account & usage

Account required
No for the listed keyless tools
Required product
Rate-limited hosted search/fetch without an account. OAuth or a team API key enables account usage and the hosted Agent tool.
Billing basis
Account usage follows the operation: search requests, returned content pages and Agent effort have different prices. Credits are shared across a team’s keys, with optional per-key budgets. Consult the current pricing table before enabling research. Provider pricing.
Server software license
MIT

When to choose Exa MCP

A useful fit when

  • An assistant that searches for source pages and fetches selected URLs.
  • Authenticated research that keeps track of a running job and its team budget.

Check another approach when

  • Assuming a search result proves every linked page was fetched successfully.
  • A raw proxy endpoint or a browser session for clicking through pages.
Sources: Exa hosted MCP setup and current tools, Exa API error handling, Exa team credits and limits

Connect to Exa MCP

Choose a connection profile below and replace credential placeholders in your own MCP client. The transport describes the MCP connection.

Search and fetch without an account

Streamable HTTP · hosted

Endpoint
https://mcp.exa.ai/mcp
Authentication
No key for rate-limited hosted search and fetching.

Before you connect

  • A client supporting remote Streamable HTTP.
  • Keep the default tool set for the simplest search-to-fetch workflow.

Cursor mcp.json (JSON) Sources: Exa hosted MCP setup and current tools, Cursor MCP configuration and headers

Cursor: .cursor/mcp.json (project) or ~/.cursor/mcp.json (global). Configuration guide.

{
  "mcpServers": {
    "exa": {
      "url": "https://mcp.exa.ai/mcp"
    }
  }
}

Sign in for team usage and Agent

Streamable HTTP · hosted

Endpoint
https://mcp.exa.ai/mcp
Authentication
OAuth prompted by the ?login URL in a compatible client.

Before you connect

  • An Exa account and an MCP client supporting OAuth.
  • Confirm the intended Exa team before account usage.

Cursor mcp.json (JSON) Sources: Exa hosted MCP setup and current tools, Exa team credits and limits, Cursor MCP configuration and headers

Cursor: .cursor/mcp.json (project) or ~/.cursor/mcp.json (global). Configuration guide.

{
  "mcpServers": {
    "exa": {
      "url": "https://mcp.exa.ai/mcp?login"
    }
  }
}

Use an Exa team API key

Streamable HTTP · hosted

Endpoint
https://mcp.exa.ai/mcp
Authentication
x-api-key header with your own Exa API key.

Before you connect

  • An Exa account/API key and client support for HTTP headers.
  • Check the key’s budget and its team’s remaining credits.

Cursor mcp.json (JSON) Sources: Exa hosted MCP setup and current tools, Exa team credits and limits, Cursor MCP configuration and headers

Cursor: .cursor/mcp.json (project) or ~/.cursor/mcp.json (global). Configuration guide.

{
  "mcpServers": {
    "exa": {
      "url": "https://mcp.exa.ai/mcp",
      "headers": {
        "x-api-key": "YOUR_EXA_API_KEY"
      }
    }
  }
}

Add advanced search and retain search/fetch

Streamable HTTP · hosted

Endpoint
https://mcp.exa.ai/mcp
Authentication
No key required for this rate-limited selected search tool set.

Before you connect

  • A client supporting remote HTTP MCP.
  • This explicit list replaces defaults and omits Agent, including after later authentication.

Cursor mcp.json (JSON) Sources: Exa hosted MCP setup and current tools, Cursor MCP configuration and headers

Cursor: .cursor/mcp.json (project) or ~/.cursor/mcp.json (global). Configuration guide.

{
  "mcpServers": {
    "exa": {
      "url": "https://mcp.exa.ai/mcp?tools=web_search_exa,web_fetch_exa,web_search_advanced_exa"
    }
  }
}

Search, inspect sources and preserve the research run

A suggested sequence using the documented tools. Check the connected server's tool schema before making a call.

  1. Check the discovered tool set

    Save a profile and reconnect. The base URL exposes search and fetch. If using the advanced profile, confirm all three listed tools are present.

    Tools: web_search_exa, web_fetch_exa, web_search_advanced_exa

    Sources: Exa hosted MCP setup and current tools, Cursor MCP configuration and headers

  2. Search before fetching only when needed

    Search for the source, select relevant URLs, then fetch them. If a URL is already known, start with fetch and inspect the returned content or URL-level failure.

    Tools: web_search_exa, web_fetch_exa

    Sources: Exa hosted MCP setup and current tools, Exa API error handling

  3. Authenticate for Agent

    Choose OAuth or the API-key profile, check team credit and key budget, and inspect Agent’s current input schema. An explicit tools list must include agent_run or the tool will stay absent.

    Tools: agent_run

    Sources: Exa hosted MCP setup and current tools, Exa team credits and limits

  4. Resume the existing run

    If Agent returns status running and an id, call agent_run with that value as runId to wait for the same job. Use previousRunId only when starting a new follow-up to completed research.

    Tools: agent_run

    Sources: Exa hosted MCP setup and current tools

Selected tools & example tasks

  • web_search_exa

    Search the web for source pages and relevant content; enabled in the default hosted tool set.

    Example task: Find current documentation for a specific framework feature, retaining source URLs.

    Sources: Exa hosted MCP setup and current tools
  • web_fetch_exa

    Fetch content from known URLs; enabled in the default hosted tool set.

    Example task: Read the selected documentation URL and inspect any returned page error.

    Sources: Exa hosted MCP setup and current tools, Exa API error handling
  • web_search_advanced_exa

    Opt-in search with additional controls and filters. An explicit tools list replaces defaults.

    Example task: Restrict research using the filters exposed in the connected tool schema.

    Sources: Exa hosted MCP setup and current tools
  • agent_run

    Authenticated hosted research. A running response returns an id that can be passed back as runId to wait for the same work.

    Example task: Start a bounded research task, then resume its existing run rather than starting a second one.

    Sources: Exa hosted MCP setup and current tools

Troubleshoot the connection

Agent or a default tool is missing

Check authentication and the tools query parameter. Authenticated hosted connections enable Agent by default, but an explicit tool list replaces that default set.

Sources: Exa hosted MCP setup and current tools

429 or incomplete fetched content

For 429, reduce request load under the applicable limits. Separately inspect URL-level errors in fetched results: a successful API request can include a failed page fetch.

Sources: Exa team credits and limits, Exa API error handling

Limits & undocumented behavior

  • Keyless hosted access offers rate-limited search and page fetching. The reviewed guide does not establish a numeric anonymous quota. Agent requires authentication.
  • A tools query parameter replaces the default tool set. A selected tool can disappear after configuration changes even when the connection itself succeeds.
  • Hosted and local package defaults can differ. The profiles and Agent availability described here apply to the hosted service.
  • No external search, page fetch or Agent task was run for this listing. Documentation does not establish target success rates or a blanket no-charge rule for failures.

Exa MCP: questions & answers

Does Exa MCP require an API key?

Hosted search and fetching can start without a key under rate limits. Agent requires account authentication; OAuth provides that without pasting an API key into this configuration.

Sources: Exa hosted MCP setup and current tools
Does enabling advanced search keep the default tools?

An explicit tools parameter replaces defaults. The advanced profile includes web_search_exa and web_fetch_exa as well as web_search_advanced_exa. Add agent_run to the list if an authenticated workflow needs it.

Sources: Exa hosted MCP setup and current tools
Does runId start another research job?

Passing the returned id as runId resumes waiting for the existing Agent run. previousRunId has a different role: it starts new follow-up research using a completed run as context.

Sources: Exa hosted MCP setup and current tools

Sources & review history

This listing is documentation-confirmed. We reviewed the primary sources below; this does not establish successful tool execution.

Record updated . Review due .

  1. Exa hosted MCP setup and current toolsChecked · Hosted authentication; Default and selected tools; Agent continuation
  2. Exa operation-specific pricingChecked · Search and content units; Agent effort pricing
  3. Exa team credits and limitsChecked · Shared team credits; Key budgets; Account rate limits
  4. Exa API error handlingChecked · 402 and 429; URL-level fetch failures
  5. Official Exa MCP repositoryChecked · Server source; MIT license
  6. Cursor MCP configuration and headersChecked · Remote MCP configuration; Project and global configuration locations

Spotted a change? Share the source and the affected setup step in Site Feedback.

Alternatives to Exa MCP

These integrations share documented capabilities with Exa MCP. Compare their setup and product requirements for your task.

Apify MCP

Find Apify Actors, inspect their inputs and pricing, run a selected Actor, and retrieve its dataset through the official MCP server.

Shared capabilities
Web search · Scraping & extraction
Connection choices
Streamable HTTP · Local stdio
Product to check
Apify Actors and platform usage
Compare setup & authentication →

Bright Data MCP

Search the public web, retrieve page content and use optional browser tools through Bright Data’s hosted or local MCP integration.

Shared capabilities
Web search · Scraping & extraction
Connection choices
Streamable HTTP · Local stdio
Product to check
Bright Data Web Unlocker API; additional products may be used by optional tools.
Compare setup & authentication →

Decodo MCP

Retrieve pages, search results and supported structured targets through Decodo’s Web Scraping API.

Shared capabilities
Web search · Scraping & extraction
Connection choices
Streamable HTTP
Product to check
Decodo Web Scraping API.
Compare setup & authentication →

Evomi MCP

Inspect Evomi proxy products, targeting and usage, rotate sessions, and run scraping jobs through a local Python MCP server.

Shared capabilities
Web search · Scraping & extraction
Connection choices
Local stdio
Product to check
Evomi Public API access plus the proxy or Scraper API product used by the tool.
Compare setup & authentication →