How to Build RevOps Workflows (with Claude Code)

TL;DR

• Three workflows, one project folder: size territories, audit account coverage, flag job changes. Each is a prompt plus a few API calls.
• It takes about 30 minutes and spends about 70 of your 1,000 free records: 50 for the coverage audit and 20 for the job-change check. Sizing costs nothing.
• Claude Code reads and writes CSV files. It does not touch your CRM unless you connect a CRM tool yourself, so the output is a file you review and import.

Prerequisites

• A LeadOcean account and API key from app.leadocean.io. The free plan has no card and 1,000 records, one-off.
• The key in your shell as $LEADOCEAN_API_KEY.
• Claude Code installed, plus curl and jq.
• A CRM account export as accounts.csv with a domain column and contacts.csv with name, title, linkedin_url. Any CRM can produce both.

export LEADOCEAN_API_KEY="paste-your-key-here"
mkdir revops && cd revops

Step 1: Give Claude Code the rules and the data

Put your territory definition in a CLAUDE.md file. Claude Code's memory docs describe it as persistent instructions read at the start of each session. Use exact filter values, not prose.

## RevOps project

### Territories
- West: hqCountry US, hqState California, Washington, Colorado
- East: hqCountry US, hqState New York, Massachusetts, Virginia
- Segment: employeeRange 51-200, industry Software Development

### Rules
- Use the leadocean MCP server. Run a free count before any call that spends records.
- Say how many records a call will spend before you make it. Wait for my go-ahead.
- Never write to the CRM. Output CSV files only.

Then add the hosted MCP server. Claude Code's docs say to run /mcp in a session to sign in to a remote server with OAuth, so no key goes into the chat.

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

/mcp

Pick leadocean, approve the sign-in and confirm it lists 13 tools. Seven of them never spend records, including both count tools.

Step 2: Size every territory for free

Count companies before you cut territories. The company count tool is free, exact and returns a breakdown by HQ state, so one call shows whether a split is balanced.

Run leadocean_count_companies for the segment in CLAUDE.md: hqCountry US,
employeeRange 51-200, industry Software Development, hasMailableContacts true.
Show the total and the top 10 HQ states. Do not call anything that spends records.

The same count over REST is GET /v1/companies/count. It takes the company-search filters.

curl -s "https://api.leadocean.io/v1/companies/count?hqCountry=US&employeeRange=51-200&industry=Software%20Development&hasMailableContacts=true" \
  -H "x-api-key: $LEADOCEAN_API_KEY" | jq '{total, state: .facets.hq_state[:5]}'

To size the buyers inside those accounts, count people. A people count carries meta.distinctCompanies, the number of employers the matched people span.

Run leadocean_count_leads with country US, employeeRange 51-200, industry
Software Development, jobLevel VP, jobFunction Sales & Business Development,
facets false. Report total, totalAll and distinctCompanies.

Step 3: Audit account coverage

Coverage is how many of your CRM accounts have at least one reachable buyer on file. Pull a few people per account for a sample of accounts, then let Claude Code mark the gaps.

GET /v1/companies/{domain}/people lists the people held at one company, one record per person returned. Keep limit small. The endpoint stops at 10,000 rows per company.

head -11 accounts.csv | tail -10 | cut -d, -f1 > sample.txt

while read -r DOMAIN; do
  curl -s "https://api.leadocean.io/v1/companies/$DOMAIN/people?limit=5" \
    -H "x-api-key: $LEADOCEAN_API_KEY" \
    | jq -c --arg d "$DOMAIN" '{domain: $d, people: .data}'
done < sample.txt > coverage.jsonl

wc -l coverage.jsonl

That is 10 accounts at up to 5 people each, so 50 records at most. A company we hold nobody for returns a 404 and costs nothing.

Read coverage.jsonl and accounts.csv. Write coverage.csv with columns:
domain, people_found, has_vp_or_above, has_work_email, gap.
Set gap to "no buyer on file" when nobody is at director level or above.
Do not guess. Leave a field empty when the record does not show it.

Step 4: Flag job changes

Stale contacts are the main source of CRM decay. Enrich returns the current role in contact_data.contact_current_experiences[] and, since October 2026, career history in resume_data. Compare the current employer with the one in your CRM.

Enrich takes exactly one of linkedin_url, email, phone or person_id. You have LinkedIn URLs in the export, so use those.

head -21 contacts.csv | tail -20 | while IFS=, read -r NAME TITLE LINKEDIN; do
  curl -s https://api.leadocean.io/v1/people/enrich \
    -H "x-api-key: $LEADOCEAN_API_KEY" -H "content-type: application/json" \
    -d "{\"linkedin_url\": \"$LINKEDIN\"}" \
    | jq -c --arg n "$NAME" --arg t "$TITLE" '{crm_name: $n, crm_title: $t, now: .data.contact_data.contact_current_experiences, past: .data.resume_data.experiences}'
done > roles.jsonl

Read roles.jsonl. For each row compare crm_title and the employer in the CRM
with the current role. Write job_changes.csv with columns: crm_name, crm_title,
current_title, current_company, status (same, changed, unknown), evidence.
Say "unknown" when the record has no current role. Past employers carry no
company_id yet, so match on company name only and flag fuzzy matches.

That is 20 records at most. Review job_changes.csv by hand, then import the changed rows through your CRM's own import tool. LeadOcean is a data source, not a CRM, so it does not sync back by itself.

Step 5: Save the workflows as reusable commands

Claude Code can run a saved prompt again next quarter. Store each step's prompt in a file and call it from a session, so a new RevOps hire repeats the same audit.

mkdir -p prompts
cat > prompts/coverage-audit.md <<'PROMPT'
Run the account coverage audit from CLAUDE.md on accounts.csv.
Count first with leadocean_count_companies and tell me the record cost.
Pull at most 5 people per account for the first 10 domains.
Write coverage.csv. Do not write to the CRM.
PROMPT

For volume, move the pull out of the chat. POST /v1/exports takes up to 50,000 rows a request and 500,000 a day. The Exports page in the app builds the same filter, picks columns, shows the record price and downloads a CSV. Bulk CSV export is not on the Free plan.

What you get

Step 2 gives real numbers. On 21 October 2026 the free company count for the example segment returned this.

{
  "total": 6143,
  "totalIsExact": true,
  "facets": {
    "hq_state": [
      { "value": "California", "count": 1681 },
      { "value": "New York", "count": 621 },
      { "value": "Texas", "count": 511 }
    ]
  },
  "credits": 0,
  "unsupportedFilters": []
}

That is 6,143 US software companies with 51-200 staff that have at least one mailable contact on file. The same day, the free people count in Step 2 (US-based VPs in Sales & Business Development, same industry and size) returned 1,649 mailable people across 1,125 employers. Without an email condition the same filters match 4,141 people.

Here is how the records add up.

The coverage file has one row per account. The row below uses placeholder values.

domain,people_found,has_vp_or_above,has_work_email,gap
acme.com,5,yes,yes,

This is the response shape from the API reference for one enriched person, with placeholder values.

{
  "success": true,
  "data": {
    "profile_data": { "profile_full_name": "Jane Doe" },
    "contact_data": {
      "has_email": true,
      "email_status": "verified",
      "contact_current_experiences": [
        { "job_title": "VP Sales", "company_name": "Acme", "company_domain": "acme.com" }
      ]
    },
    "resume_data": {
      "experiences": [
        { "job_title": "Director of Sales", "company_name": "Example Corp" }
      ]
    }
  },
  "meta": { "credits": 1 }
}

Troubleshooting

FAQ