NeuralVerge MCP Server已验证已精选

The data layer for agentic AI. One remote MCP server that gives your agent structured, real-world data in the middle of a task: web search, AI extraction of any page into typed JSON, and 340 data-source tools covering e-commerce marketplaces, official company registries, startup and funding databases, review platforms, LinkedIn and contact enrichment.

No scrapers to maintain, no five vendors to stitch together, no HTTP glue code. One URL, one sign-in, one balance.

---

What is NeuralVerge MCP?

NeuralVerge exposes its API as a remote Model Context Protocol server. Any MCP-compatible client (Claude, ChatGPT, Claude Code, Cursor, VS Code, Windsurf, Gemini CLI, MCP Inspector or your own agent) can call NeuralVerge Search, AI Extract and every Data Source directly as tools.

Each tool is a thin 1:1 wrapper around the matching NeuralVerge REST endpoint, so inputs, limits and pricing are identical to the REST API. Results come back as structured JSON inside the tool call: the model gets fields it can reason over and cite, not raw HTML or search snippets.

• Endpoint: https://mcp.neuralverge.ai/mcp
• Transport: Streamable HTTP, stateless (no session to keep alive)
• Auth: sign in with OAuth (recommended) or send an API key as Authorization: Bearer YOUR_API_KEY
• Tools: 342 (run_search, run_extract and 340 Data Source tools)
• Install: nothing to install, it is a hosted remote server

The earlier endpoint https://api.neuralverge.ai/functions/v1/mcp-server still works with an API key but does not support sign-in. Use https://mcp.neuralverge.ai/mcp for new connections.

Why agents need it

AI agents are only as good as the data they can reach. Without a data layer they rely on search snippets and open pages, and they miss what actually matters for business questions: registries, marketplaces and business databases.

Status quo	With NeuralVerge
In-house scrapers that break on every layout change	Maintained source tools that return typed JSON
Separate vendors for search, extraction and enrichment: separate contracts, keys and formats	One connection, one request format, one balance
Agents that guess or misattribute sources	Every record comes from a named source the agent can cite
Web-only research	Official registries, marketplaces and vertical databases alongside the web

Available tools

Core tools

Tool	What it does
run_search	Synchronous web search with ranked results
run_extract	Loads a page by URL and extracts structured data per natural-language instructions and/or a JSON Schema

Data Sources

Category	Tools	Coverage
E-commerce & Marketplaces	128	Amazon, eBay, Walmart, Target, Wayfair, Newegg, Shopify stores, AliExpress, Alibaba, Zalando, Rakuten Ichiba, Mercari, Yahoo! Auctions Japan, Flipkart, Trendyol, Hepsiburada, Vinted, Poshmark, leboncoin, Kleinanzeigen, OLX, Craigslist, Avito, Yandex Market
Company Registries	93	UK (Companies House), France, Spain, the Netherlands, Belgium, Luxembourg, Austria, Sweden, Norway, Denmark, Finland, Estonia, Latvia, Poland, Czech Republic, Slovakia, Slovenia, Croatia, Romania, Bulgaria, Portugal, Cyprus, Georgia, Israel, Canada, Mexico, Brazil, Argentina, Colombia, Ecuador, Panama, Australia, New Zealand, South Korea (DART), Taiwan, plus GLEIF and LEI lookup
Startup & Funding Data	93	Dealroom, PitchBook, Tracxn, Owler, Y Combinator, Product Hunt, Wellfound, Tech.eu Funding Explorer, EU-Startups, Kickstarter, Indiegogo, Republic Europe (Seedrs), WeFunder, Indie Hackers, BetaList, PitchWall, Futurepedia, There's An AI For That
Product reviews	15	Capterra, Gartner Peer Insights, Trustpilot
Social media / B2B	5	LinkedIn people search, company search, company employees, profile by name and domain, profile with email
Enrichment	5	Email enrichment, email validation, email finder, phone enrichment, US phone validation
Company intelligence	1	Crunchbase company profile

A few examples of what individual tools return:

• run_uk_companies / run_uk_company_filings: name, status, registration date, address, legal form; filing history with document links
• run_pl_companies: KRS company card with NIP/REGON, representatives, capital, shareholders, branches and history
• run_amazon_product_offers: every marketplace offer for an ASIN with price, condition, seller and delivery
• run_yahoo_auctions_items_closed_search: what items actually sold for on Yahoo! Auctions Japan
• run_techeu_rounds_search: European funding rounds filtered by country, stage, sector, minimum amount and dates
• run_gleif_companies: LEI record with legal form, registration authority, BIC/ISIN codes and direct and ultimate parents
• run_email_validation: valid / invalid / risky, catch-all, provider, confidence

Full list with descriptions: docs.neuralverge.ai/mcp-server

Quick start

1. Add the server to your client with the URL https://mcp.neuralverge.ai/mcp. In Claude: Customize → Connectors → + Add → Add custom connector. Other clients are below.
2. Sign in. The client opens a NeuralVerge sign-in page. Sign in, click Allow, and you're back in the client.
3. Ask for something that needs data, for example: "Find the registered address of Škoda Auto in the Czech company registry." The agent picks the right tool, calls it mid-task and answers from the result.

Clients that can't sign in through a browser (scripts, CI, remote shells, some IDEs) use an API key instead: create one in the NeuralVerge dashboard under Settings → API and send it as Authorization: Bearer YOUR_API_KEY.

You need a NeuralVerge account that has finished onboarding. Every call bills to your organization, exactly like the REST API.

Connect your client

Client	Sign-in	Where you add it
Claude (claude.ai, Desktop, mobile)	OAuth	Customize → Connectors
ChatGPT	OAuth	Plugins → Add custom MCP server
Claude Code	OAuth or API key	claude mcp add
Cursor	OAuth or API key	mcp.json
VS Code (GitHub Copilot)	OAuth or API key	mcp.json
Windsurf	OAuth or API key	mcp_config.json
Gemini CLI	OAuth or API key	settings.json or gemini mcp add
MCP Inspector	OAuth or API key	Web UI

Claude (claude.ai and Claude Desktop)

Open Customize → Connectors, click + Add, choose Add custom connector, name it NeuralVerge, paste https://mcp.neuralverge.ai/mcp and click Continue. If Claude asks how to identify itself, choose Register automatically and leave client ID and secret empty. Click Connect, sign in and approve access. In a chat, click + → Connectors and switch NeuralVerge on.

On Team and Enterprise plans an Owner adds the connector once under Organization settings → Connectors; members then click Connect and sign in with their own account. A connector added on claude.ai is also available in Claude Desktop and the mobile apps.

ChatGPT

Open Plugins, click + and choose Add custom MCP server. Name it NeuralVerge, enter https://mcp.neuralverge.ai/mcp under Connection, choose OAuth and leave the client credentials empty. Click Create as a plugin, sign in and approve access. In a new conversation type @ and select NeuralVerge.

Claude Code

By default the server is added for the current project only. Add --scope user to use it in every project, or --scope project to share it with your team via .mcp.json.

OAuth

claude mcp add --transport http neuralverge https://mcp.neuralverge.ai/mcp

Start claude, run /mcp, select neuralverge and follow the browser login.

API key

For CI or headless use, pass the key as a header:

claude mcp add --transport http neuralverge https://mcp.neuralverge.ai/mcp \
  --header "Authorization: Bearer YOUR_API_KEY"

Cursor

Add the server to ~/.cursor/mcp.json (all projects) or .cursor/mcp.json (one project).

OAuth

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

Open Customize in the sidebar, enable NeuralVerge, then sign in to NeuralVerge in the browser window Cursor opens and approve access.

API key

Add a headers entry and read the key from an environment variable rather than writing it into the file:

{
  "mcpServers": {
    "neuralverge": {
      "url": "https://mcp.neuralverge.ai/mcp",
      "headers": {
        "Authorization": "Bearer ${env:NEURALVERGE_API_KEY}"
      }
    }
  }
}

VS Code (GitHub Copilot)

Run MCP: Add Server → HTTP from the Command Palette, or write the entry yourself in .vscode/mcp.json (workspace) or in your user configuration (MCP: Open User Configuration).

OAuth

{
  "servers": {
    "neuralverge": {
      "type": "http",
      "url": "https://mcp.neuralverge.ai/mcp"
    }
  }
}

Run MCP: List Servers → neuralverge → Start Server and finish the NeuralVerge sign-in in the browser. If VS Code doesn't start the sign-in, use an API key.

API key

The key is prompted for once and kept out of the file:

{
  "inputs": [
    {
      "type": "promptString",
      "id": "neuralverge-key",
      "description": "NeuralVerge API key",
      "password": true
    }
  ],
  "servers": {
    "neuralverge": {
      "type": "http",
      "url": "https://mcp.neuralverge.ai/mcp",
      "headers": {
        "Authorization": "Bearer ${input:neuralverge-key}"
      }
    }
  }
}

Windsurf

In the Cascade panel click the … menu and open the MCP config file (mcp_config.json). Cascade uses at most 100 tools at a time, so turn off the ones you don't need.

OAuth

{
  "mcpServers": {
    "neuralverge": {
      "serverUrl": "https://mcp.neuralverge.ai/mcp"
    }
  }
}

Enable the server in the MCP panel and sign in to NeuralVerge when prompted.

API key

{
  "mcpServers": {
    "neuralverge": {
      "serverUrl": "https://mcp.neuralverge.ai/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

Gemini CLI

In ~/.gemini/settings.json use httpUrl for a Streamable HTTP server; a bare url is treated as SSE.

OAuth

gemini mcp add --transport http neuralverge https://mcp.neuralverge.ai/mcp

or edit ~/.gemini/settings.json yourself:

{
  "mcpServers": {
    "neuralverge": {
      "httpUrl": "https://mcp.neuralverge.ai/mcp"
    }
  }
}

Start gemini and run /mcp auth neuralverge to sign in. Gemini CLI finds the sign-in and registers itself, so no oauth block is needed. The sign-in needs a browser on the same machine.

API key

On a remote shell or in a container, add the key as a header:

{
  "mcpServers": {
    "neuralverge": {
      "httpUrl": "https://mcp.neuralverge.ai/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

Other clients

Any client that speaks Streamable HTTP works with https://mcp.neuralverge.ai/mcp. Clients that only speak SSE or stdio can connect through an MCP proxy such as mcp-remote.

OAuth

If the client supports MCP authorization, add the URL with no credentials and it will offer to sign in. When asked, choose automatic registration; there is no client ID or secret to enter.

API key

Otherwise, set the header Authorization: Bearer YOUR_API_KEY.

MCP Inspector

npx @modelcontextprotocol/inspector

Open the printed link (it carries a session token), choose the Streamable HTTP transport and enter https://mcp.neuralverge.ai/mcp. The tool list loads without credentials; calling a tool needs one of the two options below.

OAuth

Run the OAuth flow from the Inspector's authentication settings, sign in to NeuralVerge, then connect.

API key

Add the header Authorization: Bearer YOUR_API_KEY, then connect.

Raw request

curl -X POST "https://mcp.neuralverge.ai/mcp" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{
    "jsonrpc": "2.0",
    "id": 1,
    "method": "tools/call",
    "params": {
      "name": "run_email_validation",
      "arguments": { "email": "john@example.com" }
    }
  }'

Example prompts

• "Who are the directors of this UK company, and what has it filed at Companies House in the last year?"
• "Check whether this supplier is an active company in the Polish KRS and list its shareholders."
• "Show every Amazon offer for ASIN B0XXXXXXX and the cheapest seller with Prime delivery."
• "Compare the price of this product on Trendyol and Hepsiburada and list all sellers."
• "List fintech funding rounds above €10M in Germany since January."
• "Pull the Y Combinator companies from the latest batch that are hiring, with their founders."
• "Extract the pricing table from https://example.com/pricing as JSON with plan, price and limits."
• "Summarize the last 50 Trustpilot reviews of example.com by theme."

Use cases

Team	Job to be done	Tools that help
Agentic products	Ship an agent that answers with facts, not guesses	Search, Extract, any Data Source
KYC / KYB	Confirm entities, officers and owners against the registry they are filed in	Company Registries, GLEIF, LEI lookup
Due diligence	First-pass research on structure, funding and reputation	Registries, Dealroom, PitchBook, Tracxn, Owler, Trustpilot
E-commerce intelligence	Monitor prices, assortment and sellers without scrapers	24 marketplaces: products, offers, sellers, reviews
Go-to-market & outbound	Find companies and the right people inside them	LinkedIn, Crunchbase, email and phone enrichment
Market research	Map players, trends and funding	Startup & Funding Data, Product Hunt, Gartner, Capterra
Competitor analysis	Track pricing, launches, reviews and funding	Extract, Product Hunt, review platforms, funding tools
Public procurement	Read tender pages and check bidders	Extract, Company Registries
HR & recruiting	Source candidates and verify their background	LinkedIn people search, enrichment, Wellfound

Good to know

• Tool count and context. tools/list returns all 342 tools in one response (several hundred kilobytes of JSON). The model pays for these definitions in context, and clients that cap active tools will only see part of the catalogue (Windsurf, for example, allows 100 at a time). If you need just a few sources, use a client that lets you pick tools, or proxy the server and expose the subset you use.
• Listing is open, calling needs credentials. initialize and tools/list work without credentials, so a client can show the catalogue before you sign in.
• Managing access. Every client that signed in appears in the dashboard under Settings → API with a label starting with Connector:. The access does not expire on its own; revoke it there to disconnect that client immediately. Removing the server from the client alone does not revoke it.
• AI Research is REST-only. NeuralVerge AI Research (multi-step research with cross-checking and inline citations) runs asynchronously, so it is not exposed over MCP. Start research runs over the REST API.
• Errors. A call without an Authorization header fails with 401 Unauthorized and a WWW-Authenticate header that points OAuth-capable clients to sign-in. An invalid or revoked key, and errors from individual tools (invalid parameters, 402 Payment Required for usage limits or unfinished onboarding, upstream failures), are returned inside the tool result with isError: true.

Pricing

Usage is paid in points on a monthly subscription. Plans start at $50/month (1 point ≈ $0.001), with no per-source fees and no separate contracts. Search and AI Extract cost 5 points per call; Data Source tools are priced per endpoint, the same as over REST. Details: neuralverge.ai/pricing

FAQ

Do I need to install or host anything? No. It is a hosted remote server. Add https://mcp.neuralverge.ai/mcp to your client and sign in.

Do I need an API key? Not for clients that support MCP authorization: they sign you in through the browser. Use an API key from Settings → API for scripts, CI, remote machines and clients without browser sign-in.

Which clients are supported? Any client that supports MCP's Streamable HTTP transport: Claude, ChatGPT, Claude Code, Cursor, VS Code, Windsurf, Gemini CLI, MCP Inspector and custom agents built on an MCP SDK. SSE- or stdio-only clients can connect through mcp-remote.

Is the MCP server different from the REST API? No. Every tool wraps the matching REST endpoint 1:1, with the same parameters, limits and pricing, and bills to the same organization.

Are results real-time? Tools fetch from the source at call time and return the result synchronously. Long-running scrapes return partial results with partial: true rather than timing out.

Can I use only some of the sources? Yes. Pick the tools you need in your client, or proxy the server and expose a subset, to keep the model's context small.

Can I run AI Research over MCP? Not yet: research runs are asynchronous, while an MCP tool has to return its result inside the call. Use the REST API for research.

Links

• Website: neuralverge.ai
• MCP documentation: docs.neuralverge.ai/mcp-server
• Client setup: docs.neuralverge.ai/connect-a-client
• Data source catalog: neuralverge.ai/data-sources
• Pricing: neuralverge.ai/pricing