How to Generate and Qualify Leads (with Claude Code)

Claude Code finds leads well when it has a data source it can query. Give it LeadOcean over MCP, keep the ICP in a file, and it sizes, pulls and scores a list in one session.

This is the short version: real queries, a real count, no theory. It takes about 20 minutes.

TL;DR

• Connect the MCP server, count the audience for free, pull 25 people, enrich the 5 that score best.
• Time: about 20 minutes. Records spent: about 30. Counting spends none.
• The Free plan (1,000 records for life, no card) covers this run many times over.

Prerequisites

• Claude Code installed.
• A free LeadOcean account at app.leadocean.io. The MCP server signs in with OAuth, so no key is pasted.
• For the REST fallback in Step 4, an API key from the app, exported as $LEADOCEAN_API_KEY.
• A one-paragraph ICP. Ours below is a placeholder: swap in yours.

Step 1: Connect the MCP server

Add the hosted server and sign in. The command format is from the Claude Code MCP docs, September 2026.

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

Start claude and run /mcp. LeadOcean should list 13 tools. The sizing and lookup tools are free: leadocean_count_leads, leadocean_count_companies, leadocean_list_filters and leadocean_list_enum_values, plus the account and export status tools. See the LeadOcean MCP docs.

Step 2: Write the ICP into a file

Claude qualifies against what you write down, so write it down. Put this in ICP.md:

ICP
- Buyer: VP of Sales at a B2B software company, based in Canada.
- Company: 51 to 500 employees.
- Disqualify: agencies, staffing firms, anyone with "assistant" or "intern" in the title.
- Good fit signal: the person's own profile mentions outbound or pipeline.
- Output: a table with name, title, company, fit score 1 to 5, one-line reason.

Keep it to rules Claude can check against fields in the data. "Innovative companies" is not checkable. "51 to 500 employees" is.

Step 3: Size the audience before you pull anything

Ask Claude to count first. Counting is free, so there is no reason to guess.

Read ICP.md. Use leadocean_list_enum_values to resolve exact filter values,
then run leadocean_count_leads with facets=false. Tell me which filters
you used and what the number is. Do not search yet.

Claude resolves the enum strings (job level VP, job function Sales & Business Development, country CA) and counts. Enum values are exact strings, so a guess like "vp" returns nothing. The count is mailable by default: verified, catch_all_valid and catch_all addresses only.

Step 4: Pull a small batch

Search 25 people, not 2,500. A small batch tells you whether the filters are right before you spend more.

Run leadocean_search_leads with the same filters plus employeeRange
["51-200","201-500"] and excludeTitle ["assistant","intern"].
Limit 25. Then score each person against ICP.md and show the table.

The REST equivalent, for a script or CI job, is a flat JSON body with array filters. It returns 25 records and uses 25.

curl -s https://api.leadocean.io/v1/people/search \
  -H "x-api-key: $LEADOCEAN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "jobLevel": ["VP"],
    "jobFunction": ["Sales & Business Development"],
    "country": ["CA"],
    "employeeRange": ["51-200", "201-500"],
    "excludeTitle": ["assistant", "intern"],
    "limit": 25
  }'

Page with the cursor value from the previous response. One search pages to 10,000 rows, so narrow the filters rather than walking the index.

Step 5: Enrich only the winners

Search rows are thin on purpose. Take the five highest scores and enrich them by person_id. Enrich takes exactly one of linkedin_url, email, phone or person_id.

curl -s -X POST https://api.leadocean.io/v1/people/enrich \
  -H "x-api-key: $LEADOCEAN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"person_id": "PERSON_ID_FROM_SEARCH", "reveal_email": true}'

Since October 2026 the enrich response carries career history in resume_data: past roles, education, certifications and awards. Claude can use it for the second scoring pass, for example "has held a quota-carrying role before". A person with no match is a 404 and costs nothing.

In Claude Code, ask in one line.

Enrich the top 5 with leadocean_get_lead and add a column for prior roles.

What you get

This is a live run of leadocean_count_leads on 2026-10-21. Filters: jobLevel VP, jobFunction Sales & Business Development, country CA. The total is mailable (verified, catch_all_valid or catch_all).

{
  "total": 5574,
  "totalIsExact": true,
  "totalAll": 18800,
  "distinctCompanies": 3873,
  "credits": 0
}

totalAll is the same filters with no email condition. The 5,574 are the ones worth sending to. The cost of that call was zero records.

A person record has this top-level shape, from the API reference (placeholder values, not a real person):

{
  "profile_data": { "...": "name, headline, current role" },
  "contact_data": { "has_email": true, "has_phone": true, "email_status": "verified" },
  "resume_data": { "...": "past roles, education, when held" },
  "meta": { "...": "request details" }
}

Response shape from the API reference, September 2026. contact_emails appear only when reveal_email is true. Check openapi.json for the inner fields before you parse.

Claude's scored table then looks like this (placeholders):

Troubleshooting

FAQ