How to Automate Competitive Intelligence (with Claude)

TL;DR

• Claude reads each competitor's team size by function, profile, leadership and technologies through the LeadOcean connector, and a short script saves the free counts every month so you can see what moved.
• It takes about 30 minutes the first time and under 5 minutes a month after. Three competitors spend about 28 of your 1,000 free records.
• Counts and filter lookups cost 0 records. LeadOcean holds no job postings, social posts, news or web traffic, so this recipe covers the people and company side only.

Prerequisites

• A LeadOcean account from app.leadocean.io. The free plan has no card and 1,000 records, one-off.
• A Claude account that can add a custom connector. The click-by-click setup is on Connect LeadOcean to Claude.
• For the monthly script: a shell with curl and jq, and a key from the app in your environment as $LEADOCEAN_API_KEY.

export LEADOCEAN_API_KEY="paste-your-key-here"

The examples use placeholder domains: acme.com is the competitor. Swap in your own three.

Step 1: Add LeadOcean to Claude

Add the connector once and every later check takes minutes. Claude signs in with OAuth, so no key is pasted into the chat.

Name: LeadOcean
URL:  https://api.leadocean.io/mcp
Request headers: leave empty

Open a new chat, switch the connector on, and approve the sign-in. The server exposes 13 tools. Counting and lookups cost nothing: leadocean_count_leads, leadocean_count_companies, leadocean_list_filters, leadocean_list_enum_values and leadocean_get_account.

Step 2: Count each competitor's team by function

Team mix is the cheapest signal you can track. A competitor with 3 engineers and 2 salespeople in our data is a different bet from one with 80 and 40. The count tool is free, so run it as often as you like.

Use leadocean_count_leads with facets false for the domain acme.com, four times,
each time adding one jobFunction: "Engineering", "Sales & Business Development",
"Customer/Client Service", "Advertising & Marketing". Put the totals in a table.
Say what each total counts. Do not call any tool that spends records.

The same count over REST is a people search with count=true as a query parameter and limit at 1. The body is flat JSON and filters are arrays. The data array comes back empty.

curl -s "https://api.leadocean.io/v1/people/search?count=true" \
  -H "x-api-key: $LEADOCEAN_API_KEY" -H "content-type: application/json" \
  -d '{"domain": ["acme.com"], "jobFunction": ["Engineering"], "limit": 1}' \
  | jq '.meta | {total, totalIsExact}'

Treat the numbers as a sample of the people we hold, not the real headcount. They move when the dataset refreshes, which happens monthly, and when a competitor hires.

Step 3: Pull the company profile

Ask for the firmographic profile of each competitor. It costs one record per company and gives Claude the size band, industry, headquarters and the technologies found on the company website.

Use leadocean_get_company for acme.com. List employee count and band, industry,
headquarters, year founded, and funding if held. Then list the detected technologies
grouped by category. Write "not held" for any null field. Do not guess.

Some fields can be null, so "not held" is a normal answer. The REST route is POST /v1/companies/enrich.

curl -s https://api.leadocean.io/v1/companies/enrich \
  -H "x-api-key: $LEADOCEAN_API_KEY" -H "content-type: application/json" \
  -d '{"domain": "acme.com"}' \
  | jq '.data | {company: .company_data.company_name, tech: .company_detected_technologies}'

Step 4: List the leadership

Pull a short roster of senior people for each competitor. Each person returned costs one record, so a limit of 5 caps the spend at 15 for three companies. Search rows are thin: name, headline, role and email flags.

Use leadocean_search_leads with domain ["acme.com"], jobLevel ["C-Team","VP"],
excludeTitle ["assistant","intern"], limit 5. Show name, job title, function and
whether we hold an email. Do not pull more than 5 per company.

The REST call is the same search without count=true.

curl -s https://api.leadocean.io/v1/people/search \
  -H "x-api-key: $LEADOCEAN_API_KEY" -H "content-type: application/json" \
  -d '{
    "domain": ["acme.com"],
    "jobLevel": ["C-Team", "VP"],
    "excludeTitle": ["assistant", "intern"],
    "limit": 5
  }' | jq '.data[] | .profile_data | {full_name, headline}'

Need the whole roster later? GET /v1/companies/{domain}/people lists everyone we hold at a company and stops at 10,000 rows.

Step 5: Find rivals you are not tracking

New entrants hurt most when you are not watching them. Search for companies that look like your competitor on size, industry and description, then ask Claude to drop the ones you already track.

Use leadocean_search_companies with q "revenue intelligence", industry
["Software Development"], employeeRange ["51-200"], limit 10. Show name, domain,
employee band and headquarters. Remove acme.com and any domain I already track:
beta-example.com, gamma-example.com.

Each company returned is one record, so the limit caps the spend. The REST body is flat JSON.

curl -s https://api.leadocean.io/v1/companies/search \
  -H "x-api-key: $LEADOCEAN_API_KEY" -H "content-type: application/json" \
  -d '{
    "q": "revenue intelligence",
    "industry": ["Software Development"],
    "employeeRange": ["51-200"],
    "limit": 10
  }' | jq '.data[].company_data | {company_name, company_employees}'

Check the industry value first. Ask Claude to call leadocean_list_enum_values with name set to industry, because 534 values are valid and a near miss returns nothing.

Step 6: Save the counts every month

The count in Step 2 is only useful as a trend. This script runs the free counts for every competitor and function and appends them to a CSV. It spends 0 records.

#!/usr/bin/env bash
set -euo pipefail
DOMAINS=("acme.com" "beta-example.com" "gamma-example.com")
FUNCS=("Engineering" "Sales & Business Development" "Customer/Client Service" "Advertising & Marketing")
OUT=ci-snapshots.csv
[ -f "$OUT" ] || echo "date,domain,function,total" > "$OUT"
for d in "${DOMAINS[@]}"; do
  for f in "${FUNCS[@]}"; do
    body=$(jq -n --arg d "$d" --arg f "$f" '{domain:[$d], jobFunction:[$f], limit:1}')
    total=$(curl -s "https://api.leadocean.io/v1/people/search?count=true" \
      -H "x-api-key: $LEADOCEAN_API_KEY" -H "content-type: application/json" \
      -d "$body" | jq -r '.meta.total')
    echo "$(date +%F),$d,\"$f\",$total" >> "$OUT"
    sleep 0.25
  done
done

Twelve calls fit inside the free plan's 5 requests a second with the sleep 0.25 between them. Agency allows 200 a second. Run it on the first of the month with cron or any scheduler, since the dataset refreshes monthly and a weekly run mostly repeats itself.

Step 7: Have Claude write the diff

Paste the CSV into a chat and give Claude one rule: report only what the numbers show. Our data has no news or intent signals, so the brief must not invent a reason for a change.

Here is ci-snapshots.csv. Compare the two latest dates for each domain and function.
List every change of 5 people or 20 percent or more, largest first. For each, say
what the numbers show and nothing about why. End with the domains that have no
change. Do not add news, funding or product claims.

Add the Step 3 profile and Step 4 roster to the same chat when a count jumps. A new VP in a function whose count grew is a stronger signal than either alone. Then check the competitor's own site and press page, because we hold neither.

Step 8: Decide what each signal means

Raw counts need a reading rule, or the monthly diff becomes noise. Agree the rules before the first run and put them in the prompt.

• A jump in Engineering with flat Sales usually means a product push. Check the competitor's changelog next.
• A jump in Sales and Business Development with flat Engineering usually means a go-to-market push. Look at which countries the new people sit in by re-running the count with country.
• A new C-Team or VP row in the Step 4 roster is worth a note on its own, since leadership changes tend to precede a change of direction.
• A change in company_detected_technologies means they swapped a tool on their website. It says nothing about their internal stack.

Keep the rules short. Claude applies a five-line rule set the same way every month, and a long one drifts. Review the diff yourself for the first three months, then trust it for the routine changes and read only the large ones.

What you get

Step 2 is a real run. We ran it on 2026-10-21 with leadocean_count_leads, which spends 0 records, on the placeholder domain acme.com. These totals count only people with an address worth sending to (verified, catch_all_valid or catch_all), which is the tool's default.

{ "filters": { "domain": ["acme.com"] }, "total": 32, "totalIsExact": true, "totalAll": 707 }
{ "filters": { "domain": ["acme.com"], "jobFunction": ["Engineering"] }, "total": 3, "totalIsExact": true, "totalAll": 24 }
{ "filters": { "domain": ["acme.com"], "jobFunction": ["Sales & Business Development"] }, "total": 1, "totalIsExact": true, "totalAll": 35 }
{ "filters": { "domain": ["acme.com"], "jobFunction": ["Customer/Client Service"] }, "total": 0, "totalIsExact": true, "totalAll": 1 }

So on 2026-10-21 we held 32 mailable people at acme.com, 3 of them in Engineering and 1 in Sales and Business Development. totalAll is the same filters with no email condition: 707 people in all. Neither is headcount. Your numbers will differ.

The monthly file then looks like this. The rows below are the shape of the file, with the real run's date and placeholder domains.

date,domain,function,total
2026-10-21,acme.com,Engineering,3
2026-10-21,acme.com,"Sales & Business Development",1
2026-10-21,acme.com,"Customer/Client Service",0

A company profile has this shape. This is the response shape from the API reference, with placeholder values.

{
  "success": true,
  "data": {
    "company_data": { "company_name": "Acme", "company_employees": "<count>" },
    "company_detected_technologies": { "<category>": ["<technology>"] },
    "company_metrics": {},
    "meta": { "schema": "leadocean.company.v1" }
  }
}

Total spend for three competitors is 3 (profiles) plus 15 (leadership) plus 10 (discovery), so 28 records. On the free plan that is about 35 rounds. Paid plans have no per-record price: Solo is $97 a month and Agency $497, each with a fair-use pace instead of credits. See pricing.

Troubleshooting

FAQ