match_company

Resolve companies via OnFire’s Matchbox2 engine. Vector search and AI matching handle partial / fuzzy / messy names, so “Acme”, “Acme Inc.”, and “acme.com” all collapse to the same canonical identity.

When to use it

• You have a company name, website, or LinkedIn URL and need the verified LinkedIn URL.
• You’re normalising or de-duping a CRM list against a canonical identity.
• You’re about to call ai_prospecting and don’t have a company LinkedIn URL — run this first.

Skip this if the user already handed you a clean company LinkedIn URL and isn’t asking you to verify it.

Inputs

Parameter	Type	Required	Description
companies	object[]	Yes	Up to 100 company descriptors. See shape below.

Each entry can mix any of:

• names — string[]. Variants are fine: ["Acme", "Acme Inc."].
• websites — string[]. e.g. ["acme.com"].
• linkedin_urls — string[]. e.g. ["https://linkedin.com/company/acme"].

At least one of names, websites, or linkedin_urls must be provided. More signals → better match accuracy.

Returns

JSON with total_count, matched_count, and a results array in the same order as the inputs. Each matched record carries:

• matched (boolean)
• name, website, linkedin_url, linkedin_id
• size, found_employee_count
• cosine_similarity, match_type, match_reason

Example

{
  "companies": [
    { "names": ["Acme Inc"] },
    { "websites": ["techcorp.com"] },
    { "names": ["StartupCo"], "websites": ["startup.io"] }
  ]
}

Notes

• Batch up to 100 companies per call.
• Multi-signal entries (name + website) materially improve match quality — pass both whenever you have them.