Match Person

match_person (atomic)

Resolves people via Onfire's Matchbox2 engine. Vector search + AI matching, so it handles "Lana P. at Frostbyte" or just an email alias.

When to use this

• You have a person but no LinkedIn URL and need to resolve their identity (e.g. before enrichment).
• A list of CRM contacts needs LinkedIn URLs before enrichment.
• You want to verify the person at a company (e.g. confirm they still work there).

Skip this if you already have a clean person LinkedIn URL and the user isn't asking you to verify identity.

Input shape

match_person(entities=[
    {"person_full_name": "John Smith", "company_name": "Acme Inc"},
    {"person_email": "jane@techcorp.com"},
    {"person_full_name": "Lana Patel", "company_name": "Frostbyte", "job_title": "VP Eng"},
    {"person_full_name": "Alex K.", "company_name": "Northwind", "company_linkedin_url": "https://linkedin.com/company/northwind"},
])

Available fields per entry:

• person_full_name
• person_email
• company_name
• job_title (improves accuracy)
• company_linkedin_url (improves accuracy)
• company_website (improves accuracy)

Minimum: at least one of person_full_name or person_email. If you only have a name (no email), company_name is required — the matcher refuses name-only lookups because they're too ambiguous.

Batch cap: 100 entries per call.

Output shape

{
  "total_count": N,
  "matched_count": M,
  "results": [
    {
      "matched": true,
      "linkedin_url": "https://linkedin.com/in/...",
      "full_name": "John Smith",
      "job_title": "VP Sales",
      "company_name": "Acme Inc",
      "company_linkedin_url": "https://linkedin.com/company/acme",
      "company_website": "acme.com",
      "match_score": 0.91,
      "match_type": "semantic_match"
    },
    ...
  ]
}

Same-order alignment with the input. The output already includes company_linkedin_url — if you got it back, you usually don't need a separate match_company call before ai_prospecting.

Hard rules

• Batch ≤ 100 per call.
• Name without email requires company_name.
• Surface low-confidence matches (match_score low, matched: false). Ask the user to confirm rather than silently picking the top result.
• Pass every signal you have. job_title and company_linkedin_url materially improve accuracy.

Common pitfalls

• Skipping company_name on a name-only lookup. The tool will refuse — fix the input rather than retrying blindly.
• Throwing away the returned company_linkedin_url. It saves you a match_company call when you go on to ai_prospecting or other company-scoped tools.
• Conflating low match_score with "no such person". Often it just means the signals were thin. Add job_title or company website and retry.

Worked examples

Person + score in one shot.

match_person(entities=[{
    "person_full_name": "Lana Patel",
    "company_name": "Frostbyte",
    "job_title": "VP Eng"
}])

Use the returned linkedin_url for contact_data_enrichment, and the company_linkedin_url for company-scoped tools.

Bulk CRM resolve.

match_person(entities=[
    {"person_full_name": row["name"], "person_email": row["email"], "company_name": row["account"]}
    for row in csv_rows
])

Then filter matched: true rows for the next step.

Email-only.

match_person(entities=[{"person_email": "ceo@startup.io"}])

Works without name or company. Useful for inbound lead lists.

What this skill does NOT do

• It doesn't score or rank prospects — that's ai-prospecting (account-level ranking, not single-person scoring).
• It doesn't resolve companies on their own — that's match-company.
• It doesn't return emails or phones — that's contact-data-enrichment.