PitchBook vs Dealroom vs Tracxn: Which Startup Database Fits Your Agent

PitchBook vs Dealroom is usually a question about vendors, but an agent only sees the fields it gets back and what each call costs. This comparison sets the PitchBook, Dealroom and Tracxn sources side by side as NeuralVerge exposes them, over REST or MCP: what each returns, how you find a profile, and the points each call costs. It makes no claims about the platforms' own paid products, so check their current terms for those.

The short version

• —The verdict: the three overlap on "who is this company and who funded it" but return different slices. The PitchBook sources return the free public profile preview, with the valuation and funding tables paywalled. Dealroom returns a fuller company page, with valuation, revenue and headcount as ranges. Tracxn has the widest set of source pages, including a full funding history, lists, sectors, people and legal entities.
• —Who each is for: PitchBook for identifying an entity and reading its public description. Dealroom for a company page with investors by round and ownership. Tracxn for funding rounds, lists and sector context.
• —What they cost: a company or investor lookup is 5 points on all three, and a search or list is 1 point per returned record. Current plan allowances are on the pricing page.
• —MCP: every source is a tool on the hosted MCP server at https://mcp.neuralverge.ai/mcp, with the same inputs, JSON and cost as REST.

If you came for a ranking of the vendors' own products, this is not that article. It is a source-by-source view of what an agent can pull, and what that costs.

PitchBook, Dealroom and Tracxn at a glance

Each platform has its own set of source pages on NeuralVerge. The table below was generated by a script that reads the source files, so the inputs, returns and points are the ones those pages state. Figures are from the source pages, at the time of writing.

The pattern matters more than any single row. Every platform has a cheap search or list step at 1 point per record and a heavier lookup at 5 points per profile. The difference is what you can search by. PitchBook searches profile names. Dealroom searches by keyword, tier, sector or country. Tracxn's company lookups start from a profile link or id token, and its discovery routes are list pages, a unicorn search by country and a legal entity search.

A condensed view of the same pages, for the questions an agent builder usually asks first:

What each source returns

PitchBook: the public profile preview

The three PitchBook sources are a profile search, a company profile preview and an investor profile preview. They return what PitchBook shows without a subscription. For a company that is a description, website, founding year, status, employee count, industries, parent company, address, social links and FAQs. For an investor it is the description, website, founding year, status, investor types, address, social links, FAQs and counters.

The limit is stated on the source page. Sections such as valuation and funding are paywalled and come back with empty tables, and the company preview carries the number of financing rounds and the latest deal type rather than the rounds themselves. The PitchBook data guide walks through that workflow, including how to get a profile id from a name. In short, this is the cheapest way to confirm an entity exists, what it does and whether it was acquired, and it is not a source of deal history.

Dealroom: a fuller company page

The Dealroom sources are a company search, a company lookup, an investor search and an investor lookup. The company lookup returns the tagline, headquarters, founding year, sector, tier, valuation, revenue and headcount ranges, investors by round, ownership, funding events, alumni founders, footprint and the company's own investments. The investor lookup returns type, headquarters, rank, portfolio, unicorn and investment counts, biggest holdings, exits, portfolio analysis, market map, similar investors and key people.

Two details from the source pages affect an agent. Dealroom shows ranges, not exact values, without a login, so valuation, revenue and headcount come back as bounds, and the exact fields can be null. And several fields can be null or empty on a given record, so treat missing as "not shown", not as zero. The company search is the one place where a numeric valuation does appear, next to a short range label, for each matching company.

Tracxn: the widest set of source pages

Tracxn has ten source pages. The company lookup returns description, website, founding year, location, stage, sectors, total funding, latest round, valuation, investors, revenue band, employee count, competitors, legal entities, board, the three latest funding rounds and news. A separate company funding source returns every round with date, amount, round name, lead and other investors. Beyond companies there are investor, person, sector, geography and legal entity lookups, plus a company list, a unicorn search and a legal entity search.

That breadth is the reason to look at Tracxn when an agent needs context rather than a single record: a sector's totals and top companies, a country's recent rounds, or the directors on a legal entity. The trade-off is discovery. Company lookups take a profile link or id token, so the agent needs a list page, the unicorn search or a link already in hand before it can look up one company. The legal entity search defaults to India, so check the country if you work elsewhere.

One real request over REST and MCP

The same sources are reachable two ways, with the same inputs and outputs. Over REST you POST to a hyphenated endpoint such as run-dealroom-companies-search. Over MCP the hosted server at https://mcp.neuralverge.ai/mcp exposes the tool with underscores, run_dealroom_companies_search. AI research is the exception, it runs over REST only, and none of the sources in this article need it.

Here is the Dealroom company search from its source page, with a keyword and a cap of two records. The company names are real companies returned in the captured sample, not an illustration.

curl -X POST https://api.neuralverge.ai/functions/v1/run-dealroom-companies-search \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"keyword": "fintech", "max_items": 2}'

The same call as an MCP tool call:

{
  "tool": "run_dealroom_companies_search",
  "arguments": { "keyword": "fintech", "max_items": 2 }
}

The response, as captured on the source page:

{
  "session_id": "11111111-1111-1111-1111-111111111111",
  "kind": "dealroom_companies_search",
  "keyword": "fintech",
  "max_items": 2,
  "start_page": 1,
  "partial": false,
  "human": "# Dealroom — companies search\n\n- **Query:** keyword=fintech\n- **Results:** 2 of 7913 found\n- **Proxy traffic:** 17.4 KB",
  "machine": {
    "total_items": 2,
    "first_item": {
      "id": "berkshire-hathaway-corp-",
      "name": "Berkshire Hathaway",
      "url": "https://dealroom.co/companies/berkshire-hathaway-corp-/",
      "alias": "berkshire-hathaway-corp-",
      "tier": "Listed",
      "sector": "Fintech",
      "hq": "Omaha, United States",
      "country": "United States",
      "founded_year": 1839,
      "valuation": 1180000000000,
      "valuation_range": "$1.18T",
      "exit_note": "IPO 1983",
      "logo": "https://storage.googleapis.com/dealroom-images-production/2026/09/26/a1b40123091158d6c35609cf6ac1cfb5.png"
    },
    "items": [
      {
        "id": "berkshire-hathaway-corp-",
        "name": "Berkshire Hathaway",
        "url": "https://dealroom.co/companies/berkshire-hathaway-corp-/",
        "alias": "berkshire-hathaway-corp-",
        "tier": "Listed",
        "sector": "Fintech",
        "hq": "Omaha, United States",
        "country": "United States",
        "founded_year": 1839,
        "valuation": 1180000000000,
        "valuation_range": "$1.18T",
        "exit_note": "IPO 1983",
        "logo": "https://storage.googleapis.com/dealroom-images-production/2026/09/26/a1b40123091158d6c35609cf6ac1cfb5.png"
      },
      {
        "id": "jpmorgan",
        "name": "JP Morgan",
        "url": "https://dealroom.co/companies/jpmorgan/",
        "alias": "jpmorgan",
        "tier": "Listed",
        "sector": "Fintech",
        "hq": "New York City, United States",
        "country": "United States",
        "founded_year": 1799,
        "valuation": 883528000000,
        "valuation_range": "$884B",
        "exit_note": "IPO 1983",
        "logo": "https://storage.googleapis.com/dealroom-images-production/2023/01/19/9726f27cce6233556b25070db8642dd3.png"
      }
    ],
    "total_found": 7913,
    "page_count": 1,
    "blocked": false,
    "proxy_traffic_bytes": 17803
  },
  "total_points": 2
}

Two records returned, so total_points is 2. The total_found field reports 7,913 matches for the keyword, which is how you see how many more exist beyond your cap. Each record carries a numeric valuation and a short valuation_range label, plus the Dealroom link and alias you pass to the company lookup next.

The PitchBook search is shaped the same way. A keyword such as stripe with max_items set to 2 returned two company profiles and 2 points, and each carries an id that the company preview accepts. Because matching is on the profile name, the sample's first hits were Stripe Theory and Stripe Communications, which is why you read type and url before you trust a result. The PitchBook profile search page shows the full sample.

Using them together

Nothing stops an agent from combining the three. A reasonable chain for an illustrative company, Acme Oy in Finland:

1. —Search by name on PitchBook or Dealroom to find the profile, at 1 point per record.
2. —Look up the Dealroom company for the page with investors by round and ownership, at 5 points.
3. —If you need every round, look up the Tracxn company funding source for the same company, at 5 points, using a Tracxn profile link.
4. —Compare the overlapping fields yourself. The platforms use their own labels and figures, which the source pages say are returned as given.

That is 11 points for one company in the best case, before any retries. Whether the company resolves on all three is not guaranteed, so a sample of your own targets is worth running first. To connect an agent, add the MCP URL to a client and sign in through the browser. The MCP setup walkthrough for Claude shows the steps. Endpoint and parameter details change, so confirm them in the documentation.

Which route to pick depends on who decides when to call. Use REST for a scheduled job with a known list of companies, such as a weekly refresh of funding for a watchlist. Use MCP when an agent meets a company name mid-task and decides it needs the funding history or the investor's portfolio. With an agent, the model chooses the call count, so watch early runs to see how many 5-point lookups a task makes.

Cost and limits

All three platforms follow the same billing shape on the source pages. A company, investor, person, sector, geography or legal entity lookup costs 5 points, which the pages state as $5 per 1,000 lookups. A search or list costs 1 point per returned record, which is $1 per 1,000. Lookups are charged when the profile resolves, and the source pages say misses are not billed for these sources. Searches that return nothing cost nothing on the Dealroom and PitchBook pages. Points come out of the monthly allowance on your plan, and the pricing page wins over anything in a blog post.

A few worked totals, using the source-page prices:

• —A 50-company watchlist, one Dealroom company lookup each, is 250 points.
• —The same list with a Tracxn funding lookup added for each company is 500 points.
• —A 200-record Dealroom company search is 200 points, and the 60-per-page paging matters for how the runs split.
• —A PitchBook profile search of 25 records plus a preview of the three you keep is 25 plus 15, or 40 points.

There is no separate MCP rate. A call over MCP costs the same as its REST equivalent, so choosing MCP is an integration decision, not a pricing one.

On limits, the data is what each platform shows publicly or without a login, so it is not a stand-in for a paid seat. Exact PitchBook valuation and rounds are paywalled. Dealroom valuation, revenue and headcount are ranges. Throughput, latency, per-platform coverage and how often a record is refreshed are not stated on the source pages, and we have not measured them, so we make no claim about completeness or freshness. Run your own sample of companies and compare the fields against what you already hold. For each platform's own pricing and terms, check its current pages, since this article does not quote them.

How to decide

Work through these questions with five or ten real companies from your own pipeline:

• —What is the agent's actual question? "Does this entity exist and what does it do" points to the PitchBook preview. "Who invested at which round" points to Dealroom or Tracxn. "What happened across this sector or country" points to the Tracxn sector and geography sources.
• —Do you need exact numbers? Dealroom valuation, revenue and headcount come back as ranges and PitchBook valuation is paywalled. If a downstream model needs a precise figure, check which source gives it for your targets.
• —Can you find the profile? PitchBook and Dealroom have name or filter searches. Tracxn company lookups need a link or id token, so count the discovery step.
• —Where do you start? Searches at 1 point per record are cheap enough to test coverage before you pay 5 points per lookup.
• —Is the company outside the US and Europe? Coverage differs by platform and region, and the source pages do not state numbers. Test your own markets, and use the legal entity sources if registry data is part of the job.
• —Do you need people and entity data? Tracxn has person and legal entity sources. Dealroom's company page lists alumni founders and its investor page lists key people.
• —How many calls will the task make? Multiply the chain in your workflow by 5 points and by the number of companies.
• —Can you tolerate missing fields? Several fields are null or empty on a given record on each platform. Decide whether your agent treats null as unknown, and have it say so rather than guess.

For a broader look at funding data across providers, the startup funding data guide covers the other startup sources, and the Crunchbase alternative post covers that platform.

Frequently asked questions

Which is better, PitchBook or Dealroom?

They return different things through NeuralVerge, so neither is better in general. The PitchBook sources return the free public profile preview of a company or investor, with valuation and funding tables paywalled. The Dealroom sources return a fuller company page with ranges for valuation, revenue and headcount, plus investors by round. Pick by the fields your agent needs, then test a few real companies.

Can I get PitchBook funding rounds through NeuralVerge?

Not the rounds table. The PitchBook company preview carries the count of financing rounds and the latest deal type, while the valuation and funding section comes back flagged as paywalled with an empty table. For full round history, the Tracxn company funding source returns every round it lists, and the Dealroom company source returns funding events.

What does each lookup cost in points?

A company or investor lookup costs 5 points on all three platforms, which the source pages state as 5 dollars per 1,000 lookups. Searches and lists cost 1 point per returned record. Lookups are charged when the profile resolves, and the Dealroom and PitchBook searches that return nothing cost nothing. Points come out of your plan's monthly allowance.

How do I find the right profile before a lookup?

On PitchBook, search by keyword to get a profile id. On Dealroom, search by keyword, tier, sector or country, or pass a slug if you know it. On Tracxn, company lookups take a profile link or id token, so the discovery step is a list page, the unicorn search, or a link you already hold.

Can an AI agent use these sources over MCP?

Yes. Every PitchBook, Dealroom and Tracxn source is a tool on NeuralVerge's hosted MCP server at https://mcp.neuralverge.ai/mcp, for example run_dealroom_companies_search or run_pitchbook_profiles_search. Add the URL to an MCP client, sign in and click Allow. The tools take the same inputs, return the same JSON and cost the same points as the REST endpoints.

Does NeuralVerge replace a PitchBook, Dealroom or Tracxn subscription?

No. These are sources that return what the platforms show publicly or without a login, as JSON, billed per call. They do not cover everything a paid seat does, and this article makes no claim about the platforms' own products. If you need data that sits behind a login or a paywall, check what a subscription includes.

Do I need all three?

Usually not. Many agents start with one, then add a second for the gap. A common pairing is a search on one platform to find the company, then a funding lookup on another. Check that the company resolves on both before you depend on the pair.