Company identity reconciliation

Access and scope

Set CRAWLORA_API_KEY to your Crawlora key. Prefer the listed Crawlora MCP tools when connected; otherwise use the bundled scripts/crawlora.sh, which sends x-api-key to https://api.crawlora.net/api/v1. Read reference/endpoints.md for exact methods and parameters. Check application code as well as HTTP status; payloads are inside data. Stop on authentication errors, back off on 429, and retry a transient upstream failure once. Bound calls/pages by the requested scope and credit budget. Retain source IDs, URLs, source dates, and observation/crawl times separately.

Produce an inspectable mapping for supplied company/brand records or an explicitly bounded discovery set. Define whether the canonical entity means legal registrant, operating company, brand, employer, branch or parent before joining sources.

Resolve candidate identities with corroborating evidence

• Preserve original names, URLs, domains, addresses/country, supplied IDs and source dates. Normalize case/spacing/domain format for candidate matching, but retain aliases, subdomains and distinctions that can identify different branches, legal entities, divisions or products. Name similarity is not a merge rule.
• Use brand extraction for homepage/schema/social/legal-link leads. Its title/logo reflects site branding, not a corporate register. maxSpeed=true skips some identity evidence; maxAgeMs can return cached observations, so record source and freshness rather than calling every response current.
• SEC search/submissions can corroborate registrant CIK, ticker/name/aliases and filing identity; a brand need not have its own CIK, and absence from SEC is not absence of a business. Tickers can be ambiguous or change; preserve source/date.
• Employer datasets/ATS company search and a supplied/verified LinkedIn company ID offer other role-specific identifiers. Do not convert a URL slug into an invented numeric/company ID. A careers host or parent profile can serve several subsidiaries; shared infrastructure does not establish one legal entity.
• Read actually discovered official about/legal/filing pages where needed, using flat web_scrape options. Treat their relationship wording as an attributed claim with date/scope. A site footer, identical logo, shared address or linked social profile alone does not prove current ownership or an acquisition outcome.

scripts/crawlora.sh /sec/company/search q="Microsoft"
scripts/crawlora.sh /datasets/jobs/companies q="Microsoft" page=1 page_size=5
# Use supplied/verified domains and IDs for targeted brand/profile lookups.

Build a reversible mapping rather than forcing duplicates together

For each candidate join retain both source records, source-specific IDs, evidence, observed relationship, date, and decision: confirmed same scoped entity, related but distinct, unresolved, or contradictory. A parent/brand relationship belongs in a relationship field, not a silent ID collapse. Public company/person names and similarly named regional businesses need careful disambiguation.

Use multiple corroborating identity signals or an explicit authoritative relationship statement, preserving source limitations. Keep histories/rebrands separate from current equivalence; a rename does not resolve all subsidiary, country or product records. Confidence should reflect actual evidence, not name similarity rendered as an unexplained percentage. Retain original rows and a change ledger so a mapping can be reviewed and undone.

Return the scoped canonical-record table, alias/ID crosswalk, relationship and evidence ledger, conflicts and missing lookups. Do not infer legal ownership, solvency, endorsement or legitimacy beyond sources. Producing a mapping does not authorise overwriting a CRM/database, contacting companies or publishing it.