Nimble MCP Server

Name: Nimble MCP Server
Author: nimble

The Nimble MCP Server gives AI agents the ability to search, extract, map, crawl, and structure data from any website in real time.

От сообществаРаботаетНужен API-ключГлобальныйБесплатноТолько чтение

Что умеет

Nimble Crawl Run: Start a web crawl to extract content from multiple pages on a website. The crawl discovers and visits pages starting from the given URL, following links up to the specified limit. Re
Nimble Crawl Status: Check the status and progress of a running or completed crawl job. Args: crawl_id: The unique crawl identifier returned by nimble_crawl_run. tool_reasoning: Optional reasoning abo
Nimble Crawl List: List crawl jobs, optionally filtered by status. Args: status: Optional filter by crawl status: - "queued": Waiting to start - "running": Currently in progress - "succeeded": Complet

Какие данные видит

Нужен ли аккаунт

Нужен API-ключ из настроек сервиса

The Nimble MCP Server gives AI agents the ability to search, extract, map, crawl, and structure data from any website in real time. It exposes Nimble’s full web data platform as MCP tools that any compatible AI client can use.

Список инструментов сервера (18)

Технические названия из tools/list. Нужны только разработчикам.

nimble_crawl_run	Start a web crawl to extract content from multiple pages on a website. The crawl discovers and visits pages starting from the given URL, following links up to the specified limit. Results can be retrieved via nimble_crawl_status. Args: url: The starting URL for the crawl. The crawler will discover and visit linked pages from this entry point. limit: Maximum number of pages to crawl (default: 100). Higher limits take longer but provide more comprehensive coverage. sitemap: How to use the site's sitemap for URL discovery: - "include": Use sitemap in addition to link following (default) - "skip": Ignore sitemap, only follow links - "only": Only crawl URLs found in sitemap name: Optional human-readable name for this crawl job. include_paths: Optional list of regex patterns. Only URLs matching at least one pattern will be crawled. Example: ["/blog/.", "/docs/."] exclude_paths: Optional list of regex patterns. URLs matching any pattern will be skipped. Example: ["/admin/.", "/private/."] allow_subdomains: Whether to follow links to subdomains of the starting URL. Default is False (stay on same subdomain). allow_external_links: Whether to follow links to external domains. Default is False (stay on same domain). tool_reasoning: Optional reasoning about why this tool was selected. tool_description: Optional description of the tool's purpose. Returns: dict: Crawl job details including: - crawl_id (str): Unique identifier for tracking the crawl - status (str): Current status (queued, running, succeeded, failed, canceled) - url (str): The starting URL
nimble_crawl_status	Check the status and progress of a running or completed crawl job. Args: crawl_id: The unique crawl identifier returned by nimble_crawl_run. tool_reasoning: Optional reasoning about why this tool was selected. tool_description: Optional description of the tool's purpose. Returns: dict: Crawl status details including: - crawl_id (str): The crawl identifier - status (str): Current status (queued, running, succeeded, failed, canceled) - total (int): Total pages discovered - completed (int): Pages successfully crawled - failed (int): Pages that failed - pending (int): Pages still to crawl - tasks (list): Individual page task details
nimble_crawl_list	List crawl jobs, optionally filtered by status. Args: status: Optional filter by crawl status: - "queued": Waiting to start - "running": Currently in progress - "succeeded": Completed successfully - "failed": Completed with errors - "canceled": Manually terminated - "all": Show all statuses (default if not specified) limit: Maximum number of crawl jobs to return (default: 20). tool_reasoning: Optional reasoning about why this tool was selected. tool_description: Optional description of the tool's purpose. Returns: dict: List of crawl jobs with their status and metadata.
nimble_crawl_terminate	Cancel a running or queued crawl job. Args: crawl_id: The unique crawl identifier returned by nimble_crawl_run. tool_reasoning: Optional reasoning about why this tool was selected. tool_description: Optional description of the tool's purpose. Returns: dict: Confirmation of the termination with updated crawl status.
nimble_map	Discover all URLs on a website by crawling its pages and sitemap. Returns a flat list of URLs found on the site. Useful for understanding site structure before targeted extraction or crawling. Args: url: The website URL to map. The tool will discover all linked pages starting from this URL. limit: Maximum number of URLs to return. If not specified, the API default is used. sitemap: How to use the site's sitemap for URL discovery: - "include": Use sitemap in addition to link following (default) - "skip": Ignore sitemap, only follow links - "only": Only return URLs found in sitemap domain_filter: Control URL scope for discovery: - "domain": Only URLs on the exact domain (default) - "subdomain": Include URLs from subdomains - "all": Include all discovered URLs regardless of domain tool_reasoning: Optional reasoning about why this tool was selected. tool_description: Optional description of the tool's purpose. Returns: dict: A dictionary containing: - links (list): List of discovered URLs with optional title and description - success (bool): Whether the mapping completed successfully - task_id (str): Unique task identifier
nimble_task_results	Get the status and results of an async task. Use this to retrieve results from nimble_extract_async or nimble_agent_run_async. If the task is still pending, returns the current state so you can poll again later. Args: task_id: The task ID returned by nimble_extract_async or nimble_agent_run_async. tool_reasoning: Optional reasoning about why this tool was selected. tool_description: Optional description of the tool's purpose. Returns: dict: When task is pending: - task.id (str): Task ID - task.state (str): "pending" When task is complete: - The full extraction or agent run results
nimble_agent_run_async	Start an asynchronous agent run. Returns immediately with a task ID. Use this for long-running agent extractions or batch processing. Retrieve results later with nimble_task_results using the returned task ID. Args: agent: Agent name to execute. Get available agents from nimble_agents_list. params: Runtime input parameters matching the agent's input_schema. Example: {"url": "https://www.amazon.com/dp/B09V3KXJPB"} callback_url: Optional webhook URL to receive results when the agent completes. tool_reasoning: Optional reasoning about why this tool was selected. tool_description: Optional description of the tool's purpose. Returns: dict: Task details including: - task (dict): Task info with id, state, and status_url
nimble_search	Search the web using Nimble's Search API with configurable content richness. This tool provides comprehensive web search capabilities across multiple sources and formats. Use for research, fact-checking, finding sources, or gathering current information. Args: query: The search query to execute. Can be natural language questions or keywords. max_results: Maximum number of search results to return (1-100). Default is 3. Actual count may be less depending on available results. - Use 3-5 for quick lookups or simple questions - Use 10-20 for standard research tasks - Use 50+ for comprehensive research or when you need diverse sources focus: Search focus mode that determines content sources and ranking. Can be a single mode string or a list of explicit subagent names. Single mode options: - "general": Standard web search (default) - "news": Real-time news search - "location": Location-based search - "shopping": E-commerce and product search (AI-powered) - "geo": Generative engine optimization (AI-powered) - "social": Social media content search (AI-powered) Or a list of explicit agent names (e.g., ["amazon_serp", "target_serp"]) for fine-grained control over AI-powered searches. Max 10 items. output_format: Content output format for extracted pages: - "markdown": Markdown-formatted content with structure (default) - "plain_text": Plain text without formatting - "simplified_html": Clean HTML without scripts/styles locale: Language/locale code (e.g., "en", "fr", "de"). Defaults to "en". country: Country code for geo-targeted results (e.g., "US", "GB", "IL"). Defaults to "US". search_depth: Content richness level. Default is "lite". - "lite": Metadata only (title, description, URL). Fast (1-3 seconds). Use for quick lookups, getting links, or simple fact-checking. - "deep": Full page scraping with complete article text. Slower (5-15 seconds). Use for in-depth research, analysis, and detailed answers. include_answer: Generate an LLM-powered answer summary based on search results. Default is False. include_domains: Whitelist of domains to include in search results (max 50). Only results from these domains will be returned. Example: ["wikipedia.org", "britannica.com"] exclude_domains: Blacklist of domains to exclude from search results (max 50). Results from these domains will be filtered out. Example: ["pinterest.com", "facebook.com"] start_date: Filter results after this date. Format: YYYY-MM-DD or YYYY. Cannot be combined with time_range. end_date: Filter results before this date. Format: YYYY-MM-DD or YYYY. Cannot be combined with time_range. time_range: Filter by recency with predefined periods. Cannot be combined with start_date/end_date. Options: "hour", "day", "week", "month", "year" content_type: Filter by content type (only supported with focus="general"). Supports semantic groups ("documents", "spreadsheets", "presentations") and specific formats ("pdf", "docx", "xlsx", etc.). max_subagents: Maximum number of parallel agents for AI-powered focus modes (1-10). Default is 3. Only applies to shopping, social, geo focus modes and explicit agent lists. Ignored for general, news, location modes. tool_reasoning: Optional reasoning about why this tool was selected. tool_description: Optional description of the tool's purpose. Returns: dict: A dictionary containing: - results (list): List of search result objects with content and metadata - answer (str, optional): LLM-generated answer if include_answer=True
nimble_extract	Extract and parse content from a specific URL using Nimble's Extract API. This tool fetches and extracts clean, structured content from any web page. Use when you have a specific URL and need its full content. Args: url: The URL to extract content from. Must be a valid, accessible web page. output_format: Content output format for the extracted page: - "markdown": Markdown-formatted content with structure (default) - "plain_text": Plain text without formatting - "simplified_html": Clean HTML without scripts/styles locale: Language/locale code (e.g., "en", "fr", "de"). Defaults to "en". country: Country code for geo-targeted results (e.g., "US", "GB", "IL"). Defaults to "US". driver: Optional browser driver type for web extraction. If not provided, the backend will intelligently select the appropriate driver. Options: "vx6", "vx10", etc. wait: Optional wait time in milliseconds before extracting content. Useful for JavaScript-rendered pages that need time to load. Example: wait=3000 for pages that render content dynamically. tool_reasoning: Optional reasoning about why this tool was selected. tool_description: Optional description of the tool's purpose. Returns: dict: A dictionary containing: - content (str): The extracted page content in the requested format - url (str): The extracted URL - metadata (dict): Additional metadata (locale, country, driver)
nimble_extract_async	Start an asynchronous URL extraction. Returns immediately with a task ID. Use this for batch processing or when you don't need results immediately. Retrieve results later with nimble_task_results using the returned task ID. Args: url: The URL to extract content from. output_format: Content output format: - "markdown": Markdown-formatted content (default) - "plain_text": Plain text without formatting - "simplified_html": Clean HTML without scripts/styles locale: Language/locale code (e.g., "en", "fr", "de"). Defaults to "en". country: Country code for geo-targeted results (e.g., "US", "GB", "IL"). Defaults to "US". driver: Optional browser driver type ("vx6", "vx10", etc.). wait: Optional wait time in milliseconds for JavaScript-rendered pages. callback_url: Optional webhook URL to receive results when extraction completes. tool_reasoning: Optional reasoning about why this tool was selected. tool_description: Optional description of the tool's purpose. Returns: dict: Task details including: - task.id (str): Task ID to use with nimble_task_results - task.state (str): "pending", "success", or "error" - task.status_url (str): URL to check task status
nimble_agents_list	Browse the catalog of pre-built Nimble agents. Use this tool as the first step to find an existing agent for a data collection task. Each agent is purpose-built for a specific website or data type (e.g. Amazon products, LinkedIn profiles, Google Maps reviews). Typical workflow: list → get (inspect details) → run (collect data). Args: skip: Offset for pagination. Use 0 for the first page. limit: Maximum number of agents to return (1-100). query: Short keyword filter (e.g. 'amazon', 'reviews'). Use 1-2 words. Returns: Paginated list of agents with name, description, and pagination metadata.
nimble_agents_get	Get full details of a specific agent including its input/output schema. Use after nimble_agents_list to inspect an agent before running it. The response includes the agent's input_schema (what parameters it needs) and output_schema (what fields it returns), along with description and metadata. Args: agent_id: Exact agent name from nimble_agents_list or nimble_agents_publish. Returns: Agent details including name, description, input_schema, and output_schema.
nimble_agents_generate	Start creating a custom agent (initial creation only). Use this tool ONLY for the first call that begins agent creation. After this call, use nimble_agents_update_session for ALL follow-ups: answering clarification questions, retrying errors, or refining the agent. Workflow: generate (once) -> poll nimble_agents_status -> follow up via nimble_agents_update_session -> run -> publish. Status flow: - "waiting": Backend needs more info — relay the message to the user, then call nimble_agents_update_session with same session_id and their answer as prompt. - "processing": Generation in progress — use nimble_agents_status to poll. Do NOT call this tool for status checks. - "complete": Agent ready — agent_name is set, proceed to run/publish. - "error"/"failed": Something went wrong — call nimble_agents_update_session with same session_id and a fix prompt referencing the error.
nimble_agents_status	Check the current status of a generate or update session. Use this after nimble_agents_generate, nimble_agents_update_from_agent, or nimble_agents_update_session returns "processing" status. Works for both generate and update sessions — they share the same polling mechanism. Status flow: - "processing": Still in progress — call again after ~30 seconds. - "waiting": Backend needs more info — call nimble_agents_update_session with the same session_id and the user's answer as prompt. - "complete": Agent ready — agent_name is set, proceed to nimble_agents_run. - "error"/"failed": Something went wrong — check error and message fields for diagnostics. Use these to compose a retry prompt (see retry-with-fix protocol).
nimble_agents_run	Execute an agent against a target URL or set of parameters. Returns structured data collected from the target page. The params dict must match the agent's input_schema — use nimble_agents_get to inspect required fields. Typical workflow: list → get (check input_schema) → run with params. Args: agent_name: Agent name from list, get, or generate results. params: Input parameters matching the agent's input_schema. Returns: Records in data.results, plus source URL and agent name. Possible issues if no results are returned: - The target URL may be unreachable or behind a login wall. - The page structure may have changed since the agent was created. - Required params may be missing or malformed — check the agent's input_schema.
nimble_agents_publish	Save a generated agent so it becomes reusable and searchable. Call after a successful nimble_agents_generate + nimble_agents_run cycle. The agent is persisted and appears in future nimble_agents_list searches. Only publish with explicit user confirmation. Uses the same session_id from generate. Args: session_id: Session ID from the nimble_agents_generate flow. Returns: Published agent details (same shape as nimble_agents_get).
nimble_agents_update_from_agent	Create an update session from an existing agent (forking when needed).
nimble_agents_update_session	Continue refinement for an existing update/generate session.