Enrich Company with MCP

Ask Claude or Cursor about a company by domain and the agent calls get_company for you. One URL, OAuth sign-in, 1 record per lookup.

This page is about setup and prompts. For the REST endpoint and the full field list, see Enrich Company. Other clients are on the MCP hub.

Connect the server

The server is https://api.leadocean.io/mcp. It signs in with OAuth 2.1, so there is no key to paste. Create a free account first.

Claude Code. The command format is from the Claude Code MCP docs, September 2026:

claude mcp add --transport http leadocean https://api.leadocean.io/mcp

Then run /mcp in a session and finish the browser sign-in. claude mcp login leadocean does the same from the shell.

Cursor. Add a remote server to mcp.json, in the url form shown in the Cursor MCP docs, September 2026:

{
  "mcpServers": {
    "leadocean": {
      "url": "https://api.leadocean.io/mcp"
    }
  }
}

Cursor starts the OAuth sign-in the first time the server is used. Other clients follow the same pattern, see the LeadOcean MCP docs.

The tool

The tool is get_company. The server lists it as leadocean_get_company. It costs 1 record per call.

Send one of the two. Name-only lookups do not work, so the agent asks for a domain. The REST twin is POST /v1/companies/enrich, and it is the same lookup.

Use the domain when you have it. A LinkedIn URL works, but a domain is the preferred key and the less ambiguous one.

Prompts that work

Put the record limit in the prompt. "Stop after 25 records" is enough for the agent to hold a batch.

The agent picks the tool from your wording. A question about size or reach goes to count_leads, which is free. A request for firmographics on a named company goes to get_company, which costs 1 record.

What the agent gets back

The result is one leadocean.company.v1 object plus a meta block. Field groups are from the company docs, September 2026. Values below are placeholders.

{
  "company": {
    "company_data": {
      "company_name": "Example Inc",
      "company_domain": "example.com",
      "company_employees": { "number_of_employees": 1200, "number_of_employees_code": "1001-5000" },
      "company_linkedin_followers": 50000,
      "company_primary_address": { "city": "City", "country_code": "US" }
    },
    "company_detected_technologies": [
      { "technology_product": { "name": "Product" }, "technology_category": { "name2": "Category" } }
    ],
    "company_metrics": { "completion_score": 90, "marketability_score": 80 }
  },
  "meta": { "source": "...", "schema": "leadocean.company.v1" }
}

Unknown values are null, never guessed. Revenue and funding are thinly held, so an empty field is normal, not an error. Tell the agent to write "not held" in the table rather than fill the gap.

Free checks first

count_leads, list_enum_values and get_account spend no records. Run them before any paid call.

We ran count_leads with domain: ["stripe.com"] on 2026-09-30. It returned total: 10301, totalIsExact: true, credits: 0. That is mailable people at that domain (verified, catch_all_valid or catch_all), not everyone we hold. Sizing the people first tells the agent whether a company is worth a paid lookup.

Troubleshooting

FAQ