Developers / MCP · Pre-launch

Veridion MCP

Drop this page into your LLM and it knows what to do. The page is the instruction set (tool catalog, compound flows, best practices) an agent reads before its first call.

Request early access

Pre-launch.

MCP service is pre-launch. Tool names, the npm package, and config keys below are draft and they lock with service launch.

Requirements

An active Veridion subscription with API access.
A Veridion API token. The token is the single credential, no separate MCP credential.
An MCP-compatible client (Claude Desktop / Claude Code / Cursor / IBM watsonx Orchestrate / any custom agent on an MCP SDK).
Network egress to data.veridion.com.

Get API token See plans

Setup

Same JSON for Claude Desktop (claude_desktop_config.json), Claude Code (~/.claude/mcp.json), and Cursor. Restart the client and the Veridion tools appear under the veridion namespace.

mcp.jsonjson

{
  "mcpServers": {
    "veridion": {
      "command": "npx",
      "args": ["-y", "@veridion/mcp-server"],
      "env": {
        "VERIDION_ACCESS_TOKEN": "<your-access-token>"
      }
    }
  }
}

What it unlocks

Chain calls across APIs, no glue code.

Compound flows an agent runs in a single session. This is the reason to reach for the MCP; single-tool calls live on the per-API docs.

Supplier onboarding risk screen

InputSupplier name + website from an ERP row.

Chain

match_company
get_corporate_hierarchy
get_esg_score
get_esg_news
get_company_locations

OutputGreen / yellow / red verdict with cited evidence per risk.

Prospecting with a sustainability gate

InputICP in plain language: mid-market DACH manufacturers, NAICS 3339 / 3363, ESG ≥ 70.

Chain

define_market
count_companies_in_market
list_companies_in_market
get_esg_score
enrich_from_website

OutputRanked enriched account list, pre-gated on sustainability.

KYB with beneficial-ownership expansion

InputCounterparty name.

Chain

match_company
get_corporate_hierarchy
match_company
get_esg_news

OutputOwnership graph with per-node risk + adverse-media hits.

Tier-N supply-chain discovery

InputTier-1 supplier + product category.

Chain

match_company
get_corporate_hierarchy
get_company_locations
find_suppliers_by_product

OutputSupply-chain tree rooted at the Tier-1 walking outward.

M&A target pipeline

InputThesis: NAICS + revenue band + geo + exclusions.

Chain

define_market
list_companies_in_market
resolve_group
get_esg_score
enrich_from_website

OutputPre-screened target list with diligence snapshots.

Market Intelligence Command Center

InputOne company name or industry.

Chain

find_lookalike_companies
count_companies_in_market
market_revenue
industry_subsegment_mapping
list_companies_in_market

OutputFull market brief + actionable prospect list (200–1,000 rows).

Tool catalog

Every tool the MCP exposes.

Grouped by the underlying API. Signatures are draft. The locked server manifest supersedes them at launch.

Identity & enrichment

Match & Enrich V6 surface. Anchor threading via veridion_org_id.

enrich_from_website{ domain | veridion_id } → full firmographic profile

Canonical entry point. Downstream tools accept the returned veridion_org_id.

match_company{ company_name, website?, phone?, address?, registry_id? } → veridion_org_id + confidence

Requires at least one non-name identifier. Website > registry+address > precise street > phone.

get_social_profiles{ domain | veridion_id } → { linkedin, facebook, twitter, instagram, youtube, ios_app, android_app }

get_technology_stack{ domain | veridion_id } → { technologies_detected, cms }

Site-detected tech signals.

get_technographics{ domain | veridion_id } → { operational_technographics }

Broader than tech stack, covers operational tech, not just website.

get_business_activity_breakdown{ domain | veridion_id } → [{ naics_code, percent_of_activity }]

Multi-NAICS breakdown. Useful for diversified conglomerates.

get_historic_revenue{ domain | veridion_id } → [{ year, revenue }]

get_historic_employee_count{ domain | veridion_id } → [{ year, employee_count }]

get_hq{ domain | veridion_id } → HQ location block

Skinny wrapper isolating HQ from the full locations list.

Discovery

Search API V4 surface. Saved markets, filter trees, ranked results.

define_market{ industry | naics[], locations[] | market_id } → market_id, market_stats

Markets are saved filters, reusable across count / revenue / list.

find_lookalike_companies{ domain | veridion_id, limit } → [{ veridion_id, skinny_profile }], market_id

find_suppliers_by_product{ product_url | product_details, limit } → [{ veridion_id, skinny_profile }]

Reverse-index from a product back to manufacturers and distributors.

find_new_companies_by_industry{ naics, founded_after } → market_id + list

count_companies_in_market{ market_id | domain | veridion_id | naics | industry, size_filters?, location_scope? } → { count, market_id }

list_companies_in_market{ market_id, filters?, limit } → [{ veridion_id, skinny_profile }]

Expands a saved market into an actual ranked list. Pull skinny first, full profiles for finalists.

market_revenue{ market_id | domain | veridion_id | naics | industry, filters } → { total_revenue }

industry_subsegment_mapping{ naics } → [{ business_tag, count }]

Breaks a broad NAICS into sub-segments by business tag.

growth_stage_classify{ domain | veridion_id } → { stage: early | growth | mature }

Location

Location API V1 surface. Anchor-first, requires a resolved veridion_id or group_id.

get_company_locations{ veridion_id } → [{ location_id, type_l1/l2/l3, is_hq, address, building, naics }]

Paginated. Sites directly operated by one company.

get_group_locations{ group_id } → [{ location }]

Every site across the corporate group, rolled up.

Ownership & hierarchy

Corporate Groups API V1 surface. Pre-launch, leans on the Ownership data package until the API ships.

get_corporate_hierarchy{ domain | veridion_id } → { ultimate_parent, majority_owner, divisions[], affiliates[] }

Four explicit layers. Affiliates never silently roll into the corporate tree.

get_group_members{ group_id } → [{ member, relationship_role }]

resolve_group{ identifiers } → group_id

ESG

ESG API V1 surface. Wrappers hide submit + poll behind a single call.

get_esg_score{ domain | veridion_id } → { overall, pillar_scores: { E, S, G }, theme_scores[] }

get_esg_news{ domain | veridion_id, date_range? } → [{ headline, source, severity, pillars_affected }]

Adverse-media + sustainability news tied to the company.

get_sustainability_commitments{ domain | veridion_id } → [{ commitment_type, target_year, source_url }]

Products & services

Offerings catalog extracted from company sites and technographics.

get_products_and_services{ domain | veridion_id } → [{ product_name, product_url, category }]

Lists

Session-durable company sets. The only stateful surface in the MCP.

list_create{ name, companies: [{ domain | veridion_id }] } → list_id, stats

list_modify{ list_id, add[], remove[] } → status

list_get{ list_id } → full profiles

Monitoring

Change-event stream for a saved market or list.

monitor_market_changes{ market_id, data_points_to_track[] } → [{ veridion_id, changed_field, new_value, detected_at }]

Natural-language Q&A

Fallback for questions that do not cleanly map to a structured field.

ask_about_company{ veridion_id, question } → unstructured answer

Use after structured tools have resolved identity and retrieved context.

Single-tool use cases

For each intent, one tool.

Who is this company?enrich_from_website

What do they sell?get_products_and_services

Find me companies like X.find_lookalike_companies

How big is the market for Y in Germany?define_market→ count_companies_in_market | market_revenue

Where are their offices?get_company_locations

Who owns them?get_corporate_hierarchy

What's their ESG score?get_esg_score

What tech do they run?get_technology_stack

Best practices

Ten rules that change cost, latency, and hallucination rates.

Resolve once, thread the ID. Downstream calls accept veridion_org_id / group_id / division_id. Re-resolving wastes calls and risks drift.

Prefer Search for sets, Match for singles. Do not use lookalike discovery as a substitute for filtered search.

Start broad, filter down, enrich at the end. Typical funnel: 10,000 → 200 → 20 → 5.

Require at least one non-name identifier on match_company. Name alone is not sufficient.

Use the website URL as the strongest identifier. Website > registry_id + address > precise street address > phone > name.

Anchor before location. Location tools require a resolved veridion_org_id or group_id, never raw names.

Treat ESG tools as async-hidden. Batch calls; do not hot-loop.

Surface provenance. Every field carries a source and last_verified_at. Cite the source, never invent numbers.

Respect confidence. Below 0.7 on match_company, ask the user to confirm or request another identifier.

Scope the ask. Paginate locations, date-range ESG news, cap lookalike and search results. The cheapest request is the one you did not make.

Integrations

Where the MCP already plugs in.

Live integrations, native MCP clients, and orchestration platforms that can reach Veridion via the raw HTTPS APIs while native MCP support matures.

Live

IBM watsonx Orchestrate

First-party Agent Connect integration. Identifier contract and match-failure behavior carry over directly to the MCP.

Native MCP

Claude Desktop

macOS and Windows. Paste the server block into claude_desktop_config.json.

Native MCP

Claude Code

CLI. Add the server to ~/.claude/mcp.json. Tools appear under the veridion namespace.

Native MCP

Cursor

Settings → MCP → add server. Same JSON shape as Claude Desktop.

Compatible

n8n · Make · Zapier · Gumloop · Workato · SnapLogic

Call the raw Veridion HTTPS APIs directly until native MCP support lands in each platform.

Bring Veridion into your platform.

Marketplace, orchestration layer, vertical agent. If you want a first-party integration, talk to the partnerships team.

Become an integration partner

Limitations

Known gaps, stated up front.

Minimum identifier rule: company name alone will not resolve. Pair with website, registry + address, or a precise street address.
Coverage holes on very small, very new, or non-digital companies. Matches return a structured failure, not a false positive.
English-only responses today. Underlying data is global and multilingual; localized tool responses are roadmap.
Data freshness is pipeline-dependent. Firmographics weekly, ownership signals daily / profiles weekly / hierarchies monthly, locations weekly, ESG varies by module.
Corporate Groups API is in build (PM1: May–June 2026, PM2: H2 2026). Hierarchy tools lean on the Ownership data package until the dedicated API ships.
Location API V1 is in build. Batch data is shipping; the dedicated route is landing. Framework draft uses /locations/v0/ internally.
Async endpoints: ESG tools are submit + poll under the hood. High-volume batches have higher effective latency than synchronous Match / Search calls.
Rate limits are governed by the subscription plan on the API token. Plan-specific; not enumerated here.
Read-only. No create / update / write operations. Lists are the only stateful surface.
No third-party authentication on behalf of the user. All non-Veridion auth flows remain outside the MCP.

API references