How to build a cold calling tech stack (for a small agency)

A cold calling tech stack is five layers: a list, phone numbers, a dialer, a CRM and a scheduler. Most guides spend their words on the dialer. The layer that decides your connect rate is the data, so this guide builds that one with real queries and then hands off.

TL;DR

• Build the list and phone layer first with three calls: size, search, phone lookup. It takes about 30 minutes.
• A 50-person test list spends about 100 LeadOcean records at most: 50 from search, up to 50 from phone lookups. Sizing is free.
• Dialer, CRM and scheduler are your pick. They only need a CSV or a webhook from the data layer.

Prerequisites

• A LeadOcean account and key. The free plan has 1,000 records for life, no card, 5 requests a second.
• The key in your shell: export LEADOCEAN_API_KEY=...
• curl and jq.
• A dialer and a CRM that accept CSV import. Any will do.

Step 1: Pick the five layers

Write the stack down before you buy anything. Each layer has one job and one hand-off.

An agency write-up lists the same layers (BetterContact, cold calling tech stack, September 2026). LeadOcean covers the first two. It has no dialer, CRM or sequencer, so you bring those.

Step 2: Size the call list for free

Count before you spend. A count costs no records, so ask as often as you like.

Use the MCP tool leadocean_count_leads from any connected agent, or the REST search with count=true as a query parameter and limit=1:

curl -s "https://api.leadocean.io/v1/people/search?count=true&limit=1" \
  -H "x-api-key: $LEADOCEAN_API_KEY" \
  -H "content-type: application/json" \
  -d '{
    "jobLevel": ["VP", "Director"],
    "jobFunction": ["Sales & Business Development"],
    "country": ["US"],
    "industry": ["Marketing Services"],
    "hasPhone": true
  }' | jq '.meta'

The filters are arrays of exact strings. If a value comes back empty, resolve it with leadocean_list_enum_values or GET /v1/enums first.

Count it for companies too. GET /v1/companies/count is free and returns breakdowns by industry, headcount band and HQ country.

Step 3: Pull the first 50 people

Search with the same filters, limit 50. Each person returned costs one record.

curl -s "https://api.leadocean.io/v1/people/search" \
  -H "x-api-key: $LEADOCEAN_API_KEY" \
  -H "content-type: application/json" \
  -d '{
    "jobLevel": ["VP", "Director"],
    "jobFunction": ["Sales & Business Development"],
    "country": ["US"],
    "industry": ["Marketing Services"],
    "hasPhone": true,
    "limit": 50
  }' > page1.json

jq -r '.meta.nextCursor' page1.json

To get the next page, add "cursor": "<nextCursor>" to the body. One search pages to 10,000 rows. Past that, narrow the filters instead of walking further.

Want only companies on a given tool? Add technologyCategories (for example CRM) to the filter. That covers about 7.8M crawled sites.

Step 3b: Narrow by what the company runs

Filter on detected technology so the first line of the call says something true about the company.

curl -s "https://api.leadocean.io/v1/people/search?count=true&limit=1" \
  -H "x-api-key: $LEADOCEAN_API_KEY" \
  -H "content-type: application/json" \
  -d '{
    "jobLevel": ["VP", "Director"],
    "jobFunction": ["Sales & Business Development"],
    "country": ["US"],
    "technologyCategories": ["CRM"],
    "employeeRange": ["51-200", "201-500"],
    "hasPhone": true
  }' | jq '.meta'

Enum values are exact strings. employeeRange takes brackets like 51-200, and technologyCategories has 105 values. Resolve any you are unsure of with leadocean_list_enum_values.

Step 4: Fetch the phone numbers

hasPhone: true tells you a number exists. To read it, call GET /v2/people/phone once per person, with the ID from the search row (meta.source_ids.person_id) sent as contact_id.

for id in 1234567 1234568; do
  curl -s "https://api.leadocean.io/v2/people/phone?contact_id=$id" \
    -H "x-api-key: $LEADOCEAN_API_KEY" | jq '.data'
  sleep 0.25
done

This costs one record per call, including an answer of found: false. A 404 (no record for that person at all) costs nothing. The sleep keeps you under the 5 requests a second on Free. Agency allows 200.

For lists beyond a few hundred rows, skip the loop. Use POST /v1/exports (not on Free) or the Exports page in app.leadocean.io: filter builder, column picker, price preview, CSV download. An export takes up to 50,000 rows a request and 500,000 a day.

Step 5: Load the dialer and CRM

Both take a CSV. Map four columns and you are done.

Keep the person_id column. It stays stable across exports, so you can de-duplicate on re-import. Check each tool's own import docs for the exact header names. Read email_1_status before any email follow-up.

Step 6: Let an agent do the sizing

If you work in Claude Code or another MCP client, connect https://api.leadocean.io/mcp (OAuth sign-in, no key pasted). Then ask in plain words:

Count US VP and Director level sales leaders at Marketing Services companies
with a phone number on file. Use leadocean_count_leads only. Do not search.
Then repeat for employeeRange 51-200 and report both totals.

The count tools are free, so an agent can iterate on filters without spending a record. Keep search, get_lead and find_phone for the final list.

What a first week costs

Plan the spend by layer. Data is the only one LeadOcean prices here.

That fits inside the Free plan's 1,000 records ten times over. A team calling daily outgrows it fast. Agency at $497 a month has a fair-use pace instead of credits, plus exports of 50k rows a file. Solo is $97 a month for one person. See pricing for what each plan includes.

What you get

Step 2 count run on 2026-10-21 with leadocean_count_leads (free):

{
  "filters": "jobLevel VP or Director, jobFunction Sales & Business Development, country US, industry Marketing Services, hasPhone true",
  "total": 3774,
  "totalIsExact": true,
  "distinctCompanies": 1528,
  "credits": 0
}

That is a phone-flagged count, not a mailable count. It filtered on hasPhone, so it says nothing about email deliverability.

The phone call returns this shape (response shape from the API reference, placeholder values):

{
  "success": true,
  "data": {
    "found": true,
    "best": { "phone": "+1 415 555 0133", "type": "mobile" },
    "phones": [
      { "phone": "+1 415 555 0133", "type": "mobile" },
      { "phone": "+1 415 555 0134", "type": "office" }
    ]
  }
}

The array is ordered best first. Each number carries only phone and type. The response has no meta, so read spend from GET /v1/usage.

Troubleshooting

FAQ