Olostep MCP Server

Docker Hub npm version License: ISC

Live web search, scraping and crawling API for AI. Get structured, LLM-ready data from any site. Scale to 1M requests with 10 parallel batches at low cost.

Olostep is web data infrastructure for AI agents and enterprise AI workflows. Its APIs provide live web search, scraping and crawling to extract, clean and structure data from the web. This is the official Olostep Model Context Protocol (MCP) server, which gives any MCP-compatible AI agent those capabilities.

Use the hosted server at https://mcp.olostep.com/mcp (sign in with OAuth, or use an API key), or run it locally with an API key from the Olostep website.

Why Olostep

• Web data infrastructure for developers and enterprises

• Scales to hundreds of millions to billions of requests per customer per month

• The most performant, reliable and cost-effective API on the market

• Used by the fastest-growing AI companies

From extracting data from a handful of URLs to crawling hundreds of millions of pages, Olostep gives developers and AI companies reliable, scalable and highly cost-effective access to clean, structured web data.

Related MCP server: Parallel Task MCP

Features

• Scrape website content in HTML, Markdown, JSON or Plain Text (with optional parsers)

• Parser-based web search with structured results

• AI Answers with citations and optional JSON-shaped outputs

• Batch scraping of up to 100k URLs

• Autonomous site crawling from a start URL

• Website URL discovery and mapping (with include/exclude filters)

• Country-specific request routing for geo-targeted content

• Configurable wait times for JavaScript-heavy websites

• Comprehensive error handling and reporting

• Simple API key configuration

Installation

There are multiple ways to connect to the Olostep MCP Server. Choose the one that best fits your workflow.

One-click install

Connect to the hosted server and sign in with your Olostep account (OAuth). No API key or local setup needed.

Or from the terminal:

# Claude Code
claude mcp add --transport http olostep https://mcp.olostep.com/mcp

# Gemini CLI (prompts for your API key)
gemini extensions install https://github.com/olostep/olostep-mcp-server

For other clients, see Client Setup.

☁️ Remote Endpoint (Recommended)

The simplest way — no local installation required. Connect directly to our hosted MCP server:

https://mcp.olostep.com/mcp

Authentication is done via a Bearer token in the Authorization header using your Olostep API key. See the Client Setup section below for configuration examples.

🐳 Docker Hub

Pull and run the official Docker image:

docker pull olostep/mcp-server

docker run -i --rm \
  -e OLOSTEP_API_KEY="your-api-key" \
  olostep/mcp-server

🔧 Local Docker Build

If you prefer to build the image yourself from source:

git clone https://github.com/olostep/olostep-mcp-server.git
cd olostep-mcp-server
npm install
npm run build
docker build -t olostep/mcp-server:local .

docker run -i --rm -e OLOSTEP_API_KEY="your-api-key" olostep/mcp-server:local

📦 npx

Run without any installation using npx:

env OLOSTEP_API_KEY=your-api-key npx -y olostep-mcp

On Windows (PowerShell):

$env:OLOSTEP_API_KEY = "your-api-key"; npx -y olostep-mcp

On Windows (CMD):

set OLOSTEP_API_KEY=your-api-key && npx -y olostep-mcp

Or install globally:

npm install -g olostep-mcp

Client Setup

Gemini CLI

Install the Olostep extension in Gemini CLI (requires an Olostep API key):

gemini extensions install https://github.com/olostep/olostep-mcp-server

Configure your API key through Gemini CLI's sensitive extension settings:

gemini extensions config olostep OLOSTEP_API_KEY

Gemini CLI stores this value in your system keychain. Restart Gemini CLI, then verify the extension and MCP server:

gemini extensions list
gemini mcp list

For authentication errors, configure the key again. For connection errors, restart Gemini CLI and run gemini mcp list. See the official Gemini CLI extensions documentation for details.

Cursor

The easiest way is to use the remote endpoint. Create or edit .cursor/mcp.json in your project root:

{
  "mcpServers": {
    "olostep": {
      "url": "https://mcp.olostep.com/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY_HERE"
      }
    }
  }
}

Alternative (local): Go to Cursor Settings > Features > MCP Servers, click "+ Add New MCP Server":

• Name: olostep

• Type: command

• Command: env OLOSTEP_API_KEY=your-api-key npx -y olostep-mcp

Claude Desktop

Add this to your claude_desktop_config.json:

{
  "mcpServers": {
    "mcp-server-olostep": {
      "command": "npx",
      "args": ["-y", "olostep-mcp"],
      "env": {
        "OLOSTEP_API_KEY": "YOUR_API_KEY_HERE"
      }
    }
  }
}

Alternative (Docker):

{
  "mcpServers": {
    "olostep": {
      "command": "docker",
      "args": [
        "run", "-i", "--rm",
        "-e", "OLOSTEP_API_KEY=YOUR_API_KEY_HERE",
        "olostep/mcp-server"
      ]
    }
  }
}

Or install via the Smithery CLI in your device terminal:

npx -y @smithery/cli install @olostep/olostep-mcp-server --client claude

Claude Code

Add the remote endpoint to your Claude Code MCP configuration:

{
  "mcpServers": {
    "olostep": {
      "url": "https://mcp.olostep.com/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY_HERE"
      }
    }
  }
}

Alternative (local):

{
  "mcpServers": {
    "olostep": {
      "command": "npx",
      "args": ["-y", "olostep-mcp"],
      "env": {
        "OLOSTEP_API_KEY": "YOUR_API_KEY_HERE"
      }
    }
  }
}

Windsurf

Add this to your ./codeium/windsurf/model_config.json:

{
  "mcpServers": {
    "olostep": {
      "serverUrl": "https://mcp.olostep.com/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY_HERE"
      }
    }
  }
}

Alternative (local):

{
  "mcpServers": {
    "mcp-server-olostep": {
      "command": "npx",
      "args": ["-y", "olostep-mcp"],
      "env": {
        "OLOSTEP_API_KEY": "YOUR_API_KEY_HERE"
      }
    }
  }
}

Cline

Add this to your cline_mcp_settings.json (Cline → MCP Servers → Configure):

{
  "mcpServers": {
    "olostep": {
      "command": "npx",
      "args": ["-y", "olostep-mcp"],
      "env": {
        "OLOSTEP_API_KEY": "YOUR_API_KEY_HERE"
      },
      "disabled": false,
      "autoApprove": []
    }
  }
}

Cline can also set this up for you: ask it to install the MCP server from this repository, and it follows llms-install.md.

VS Code

Add this to your .vscode/mcp.json:

{
  "servers": {
    "olostep": {
      "type": "http",
      "url": "https://mcp.olostep.com/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY_HERE"
      }
    }
  }
}

Alternative (local):

{
  "servers": {
    "olostep": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "olostep-mcp"],
      "env": {
        "OLOSTEP_API_KEY": "YOUR_API_KEY_HERE"
      }
    }
  }
}

GitHub Copilot CLI

Inside Copilot CLI, run /mcp add and enter:

• Name: olostep

• Type: HTTP

• URL: https://mcp.olostep.com/mcp

• Headers: leave empty

Save with Ctrl+S, then sign in with your Olostep account when prompted (OAuth). Check the connection with /mcp show olostep.

This writes the following to ~/.copilot/mcp-config.json:

{
  "mcpServers": {
    "olostep": {
      "type": "http",
      "url": "https://mcp.olostep.com/mcp",
      "tools": ["*"]
    }
  }
}

Zed

Add this to your Zed settings.json:

{
  "context_servers": {
    "olostep": {
      "url": "https://mcp.olostep.com/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY_HERE"
      }
    }
  }
}

Alternative (extension): the Olostep Zed extension runs the local server and asks for your API key in Zed's setup screen.

Metorial

Option 1: One-Click Installation (Recommended)

1. Open Metorial dashboard

2. Navigate to MCP Servers directory

3. Search for "Olostep"

4. Click "Install" and enter your API key

Option 2: Manual Configuration

Add this to your Metorial MCP server configuration:

{
  "olostep": {
    "command": "npx",
    "args": ["-y", "olostep-mcp"],
    "env": {
      "OLOSTEP_API_KEY": "YOUR_API_KEY_HERE"
    }
  }
}

The Olostep tools will then be available in your Metorial AI chats.

Configuration

Environment Variables

• OLOSTEP_API_KEY: Your Olostep API key (required)

• ORBIT_KEY: An optional key for using Orbit to route requests.

Available Tools

1. Scrape Website (scrape_website)

Use this when you have a URL and need what is on the page. Returns markdown (default), html or text with the page's url and title, after rendering JavaScript. Long pages are cut at max_characters (default 20,000) and link to their full text.

{
  "name": "scrape_website",
  "arguments": {
    "url_to_scrape": "https://example.com",
    "output_format": "markdown",
    "country": "US",
    "wait_before_scraping": 1000,
    "parser": "@olostep/amazon-it-product"
  }
}

Parameters:

• url_to_scrape: The URL of the website you want to scrape (required)

• output_format: Choose format (html, markdown, json, or text) - default: markdown

• country: Optional country code (e.g., US, GB, CA) for location-specific scraping

• wait_before_scraping: Wait time in milliseconds before scraping (0-10000)

• parser: Optional parser ID for specialized extraction

• max_characters: Max characters of page content to return (default: 20000). Raise it to read more of a long page.

Response (example):

The page's url and title come first. When the content is longer than max_characters, it is cut and truncated gives the full length and a full_text_url. JSON output is never cut: it is returned whole up to 100,000 characters. Above that, json_content is left out and json_content_omitted gives its size and a json_url to fetch the full JSON.

{
  "url": "https://en.wikipedia.org/wiki/Web_scraping",
  "title": "Web scraping - Wikipedia",
  "markdown_content": "Web scraping - Wikipedia ...",
  "markdown_hosted_url": "https://olostep-storage.s3.us-east-1.amazonaws.com/markDown_....txt",
  "truncated": [
    { "field": "markdown_content", "returned_characters": 20000, "total_characters": 57235, "full_text_url": "https://olostep-storage.s3.us-east-1.amazonaws.com/markDown_....txt" }
  ],
  "truncation_note": "Content was cut at 20000 characters. ..."
}

2. Search the Web (search_web)

Search the web for a query and get each result's URL, title and description. Optionally get each result page's content as markdown in the same call.

{
  "name": "search_web",
  "arguments": {
    "query": "your search query",
    "limit": 10,
    "include_domains": ["docs.olostep.com"],
    "scrape_results": true
  }
}

Parameters:

• query: Search query in natural language (required)

• limit: Number of results, from 1 to 25 (default: 10)

• include_domains: Only return results from these domains (optional)

• exclude_domains: Never return results from these domains (optional)

• scrape_results: Also return each result page's content as markdown (default: false). Each scraped page uses extra credits.

• max_characters: With scrape_results, the maximum characters of content per page (default: 20000)

Response:

• JSON (as text) with query, search_id, total_results and results: each result has url, title and description, plus markdown_content when scrape_results is true. Cut pages list the cut fields under truncated.

3. Answers (AI) (answers)

Search the web and return AI-powered answers in the JSON structure you want, with sources and citations.

{
  "name": "answers",
  "arguments": {
    "task": "Who are the top 5 competitors to Acme Inc. in the EU?",
    "json": "Return a list of the top 5 competitors with name and homepage URL"
  }
}

Parameters:

• task: Question or task to answer using web data (required)

• json: Optional JSON schema/object or a short description of the desired output shape

Response includes:

• answer_id, object, task, result (JSON if provided), sources, created

4. Batch Scrape URLs (batch_scrape_urls)

Scrape up to 100k URLs in one batch. Perfect for large-scale data extraction. New accounts start at 100 URLs per batch, and info@olostep.com can lift this.

{
  "name": "batch_scrape_urls",
  "arguments": {
    "urls_to_scrape": [
      {"url": "https://example.com/a", "custom_id": "a"},
      {"url": "https://example.com/b", "custom_id": "b"}
    ],
    "output_format": "markdown",
    "country": "US",
    "wait_before_scraping": 500,
    "parser": "@olostep/amazon-it-product"
  }
}

Response includes:

• batch_id, status, total_urls, created_at, formats, country, parser, urls

5. Create Crawl (create_crawl)

Start an async crawl that autonomously discovers and scrapes entire websites by following links. Returns a crawl_id — the crawl runs in the background and does not return content in this response. You must then call get_crawl_results with the crawl_id to poll status and retrieve the scraped pages (same two-step pattern as batch_scrape_urls + get_batch_results).

{
  "name": "create_crawl",
  "arguments": {
    "start_url": "https://example.com/docs",
    "max_pages": 25,
    "output_format": "markdown",
    "country": "US",
    "parser": "@olostep/doc-parser"
  }
}

Response includes:

• crawl_id, object, status, start_url, max_pages, created, formats, country, parser

Pair this call with get_crawl_results — do not pass a crawl_id to get_batch_results (crawls and batches are separate resources).

6. Create Map (create_map)

Get all URLs on a website. Extract all URLs for discovery and analysis.

{
  "name": "create_map",
  "arguments": {
    "website_url": "https://example.com",
    "search_query": "blog",
    "top_n": 200,
    "include_url_patterns": ["/blog/**"],
    "exclude_url_patterns": ["/admin/**"]
  }
}

Response includes:

• map_id, object, url, total_urls, urls, search_query, top_n

7. Get Batch Results (get_batch_results)

Retrieve the results of a previously submitted batch scrape job using its batch_id.

{
  "name": "get_batch_results",
  "arguments": {
    "batch_id": "batch_abc123"
  }
}

Parameters:

• batch_id: The batch ID returned from batch_scrape_urls (required)

• items_limit: Max items to retrieve content for per call, 1–100 (default: 20)

• cursor: Pass next_cursor from the previous result to get the next pages (default: 0)

• max_characters: Max characters of content per URL (default: 20000). Longer pages are cut and link to their full text.

Large batches return their pages a set at a time. When has_more is true, call again with next_cursor until you have every page.

Response includes:

• batch_id, status (processing or completed), total_urls, completed_urls, items (array of scraped results per URL with url, custom_id, markdown_content, html_content, json_content, text_content, status, page_metadata)

8. Get Crawl Results (get_crawl_results)

Retrieve the status and scraped pages for an async crawl started with create_crawl. This is the required companion to create_crawl — create_crawl only kicks off the job and returns a crawl_id; this tool is how you actually fetch the discovered pages and their content.

{
  "name": "get_crawl_results",
  "arguments": {
    "crawl_id": "crawl_abc123",
    "formats": ["markdown"],
    "items_limit": 20,
    "cursor": 0
  }
}

Parameters:

• crawl_id: The crawl ID returned from create_crawl (required)

• formats: Array of formats to retrieve per page — markdown, html, json, text (default: ["markdown"])

• items_limit: Max pages to retrieve content for, 1–100 (default: 20)

• cursor: Pagination cursor into the list of discovered pages (default: 0)

• search_query: Optional filter to rank/select pages by relevance to a query

• max_characters: Max characters of content per page (default: 20000). Longer pages are cut and link to their full text.

Response includes:

• While in progress: crawl_id, status (in_progress), pages_completed, pages_total, and a message prompting you to call again in ~10 seconds.

• When completed: crawl_id, status (completed), pages_returned, next_cursor, has_more, and a pages array where each entry has url, title, custom_id, and the requested content fields (markdown_content, html_content, json_content, text_content).

Older tool names

get_webpage_content and get_website_urls duplicated scrape_website and create_map. They are no longer listed to clients, so models choose between fewer, clearer tools, but calls to them still work. Use scrape_website and create_map instead.

Error Handling

The server provides robust error handling:

• Detailed error messages for API issues

• Network error reporting

• Authentication failure handling

• Rate limit information

Example error response:

{
  "isError": true,
  "content": [
    {
      "type": "text",
      "text": "Olostep API Error: 401 Unauthorized. Details: {\"error\":\"Invalid API key\"}"
    }
  ]
}

Distribution

Docker Images

The MCP server is available as a Docker image:

• Docker Hub: [olostep/mcp-server](https://hub.docker.com/r/olostep/mcp-server)

• Official Docker MCP Registry: mcp/olostep (coming soon - enhanced security with signatures & SBOMs)

• GitHub Container Registry: ghcr.io/olostep/olostep-mcp-server

Docker Desktop MCP Toolkit

The Olostep MCP Server is being added to Docker Desktop's official MCP Toolkit, which means users will be able to:

• Discover it in Docker Desktop's MCP Toolkit UI

• Install it with one click

• Configure it visually

• Use it with any MCP-compatible client (Claude Desktop, Cursor, etc.)

Status: Submission in progress to Docker MCP Registry

Supported Platforms

• linux/amd64

• linux/arm64

Building Locally

# Clone the repository
git clone https://github.com/olostep/olostep-mcp-server.git
cd olostep-mcp-server

# Build the image
npm install
npm run build
docker build -t olostep/mcp-server .

# Run locally
docker run -i --rm -e OLOSTEP_API_KEY="your-key" olostep/mcp-server

About Olostep

Olostep is web data infrastructure for AI. It provides live web search, scraping and crawling API to extract and structure data, used in production by AI companies to give their agents fresh, LLM-ready web data at scale.

• Website: https://www.olostep.com

• Documentation: https://docs.olostep.com

• Get an API key: https://olostep.com/auth

License

ISC License