Can I resell or embed the data in my product?

Use in your product is governed by our terms of service. For redistribution, volumes and custom limits, talk to us at sales@neuralverge.ai before you launch.

How do I keep costs predictable when an agent decides what to call?

Every action has a fixed points cost listed on the pricing page, and usage comes out of one monthly balance. Limit which tools your agent can see, and cap research depth per request, to keep spend in line with what each feature is worth.

What happens when a source changes its layout?

Keeping sources working is on our side, not yours. Your agent keeps calling the same tool with the same inputs and gets the same response shape back.

Which agent frameworks does it work with?

Any client or framework that speaks MCP, and anything that can make an HTTP request for REST. There is no SDK you have to adopt.

Use case · Agentic Products

Real-world data for the agent inside your product

Your copilot, assistant or autonomous agent is only as useful as what it can look up. NeuralVerge gives it company, people, registry and marketplace data, web search and page extraction as tools — structured JSON it can reason over, not HTML it has to parse.

Get started Connect via MCP

In practice

What you ask, and what comes back

You ask

“A user in your CRM copilot asks: “Who is acme-logistics.de, how big are they, and who runs it?””

Agent runscompany profileregistry recordpeople search

Your agent answers from structured fields instead of guessing from its training data.

Field	Value	From
Legal name	Acme Logistics GmbH	Company registry
Status	Active, registered 2014	Company registry
Headcount	51–200	Company profile
Managing director	K. Brandt	Registry · public profile

Illustrative output — company and contact names are fictional.

Why it changes

Scrapers, stale training data, one integration per source

Most agent products start with a web search tool and a pile of scrapers. It works in the demo, then breaks in production.

The usual way

The model guesses from training data

Ask an LLM about a company and it answers confidently from whatever it saw during training — out of date, unsourced, and sometimes about a different company with the same name.

With NeuralVerge

The agent looks it up

Company, people, registry and marketplace lookups come back as structured fields read at request time, and research runs return answers with inline citations your users can check.

The usual way

One scraper per source, all of them yours

Each new source means another integration, another response format and another thing that breaks when a site changes its layout — maintained by your team, not your vendor.

With NeuralVerge

One request format for every source

Every source in the catalog is called the same way, so adding one to your product is a new tool name, not a new integration. Keeping sources working is our job.

The usual way

Raw HTML in the context window

A search tool returns pages; the agent burns tokens reading them and still has to pull out the fields it needed, with mixed results.

With NeuralVerge

Structured JSON your agent can reason over

Data sources return typed fields, and page extraction turns any URL into JSON that matches the schema you pass — smaller context, more predictable output.

How it works

Start in three steps

Connect

Point your agent framework at the MCP server, or call the REST endpoints from your own code. Both use the same API key and return the same response shape.

Expose the tools

Give your agent the sources it needs — every data source, search and extraction is an MCP tool. Research runs are available over REST.

Ship

Your agent calls the tools mid-task and gets structured results back. You pay per lookup from one points balance, the same over MCP or REST.

Claude Code and Claude Desktop: run this once. It registers NeuralVerge as an HTTP MCP server, and its tools appear in your next conversation. Replace YOUR_ACCESS_TOKEN with your NeuralVerge token.

claude mcp add --transport http neuralverge \
  https://api.neuralverge.ai/functions/v1/mcp-server \
  --header "Authorization: Bearer YOUR_ACCESS_TOKEN"

On repeat

Patterns that hold up in production

How teams wire NeuralVerge into an agent product — from a single tool call per user question to scheduled jobs that keep records current.

Per request

Live

The agent looks things up while it answers.

Company or person lookup in a copilotChat
Validate an email before a sendWorkflow

Per task

Async

Longer jobs run while the user does something else.

Deep research with citationsReport
Extract a list of pages to one schemaDatabase

Nightly

Batch

Records in your product stay current.

Re-check company records users followDatabase
Refresh marketplace listingsDatabase

Schedules run in your agent or your own code. NeuralVerge answers the lookups; it does not run or send anything on its own.

Data behind it

The sources this runs on

Browse all sources

Corporate registry

93 sources

Official company records — status, officers, filings — from national registries.

See the sources

Company intelligence

16 sources

Company profiles, competitors, news, and software reviews and ratings.

See the sources

Social media

7 sources

Public people and company profiles, people search and employee lists.

See the sources

E-commerce & Marketplaces

128 sources

Products, prices, offers, sellers and reviews from online marketplaces.

See the sources

FAQ

Common questions

REST fits code that already knows when a call should happen. MCP fits an agent that decides for itself, mid-task, which tool to reach for. Both use the same key and return the same response shape, and a call costs the same either way. Research runs are REST-only.

Give your agent something real to work with

Company, people, registry and marketplace data, search and extraction — as tools your agent can call, on one API key.

Get started