company-contact-finder（公司联系方式查找器）

company-contact-finder

Find decision-makers at a specific company by name and target titles. Uses Gooseworks MCP tools (Apollo, Crustdata, Fiber, PDL) with a layered fallback strategy to maximize results while minimizing cost.

Inputs

Procedure

Step 1: Understand the Request

Parse the user's request to extract:

• company_name (required) -- the company to search at
• company_linkedin_url (optional) -- helps disambiguate common names
• target_titles (required) -- list of job titles or roles to find (e.g., ["Partner", "Controller", "VP Finance", "CFO"])
• num_results (optional, default 10) -- how many contacts to return

If the user does not provide target titles, ask for them. Suggest common senior titles based on context:

• For accounting/CPA firms: Partner, Managing Director, Controller, CFO, VP Finance
• For tech companies: VP Engineering, CTO, Head of Product, Director of Engineering
• For general B2B: VP, Director, C-Level, Head of

Step 2: Apollo Search (Primary — cheapest at $0.01/call)

Apollo is the cheapest search provider. Start here for all searches.

Call:

apollo_person_search(
  person_titles: ["Partner", "Controller", "VP Finance"],
  organization_domains: ["eisneramper.com"],
  per_page: 25
)

If you don't have the company domain, use q_keywords with the company name:

apollo_person_search(
  person_titles: ["Partner", "Controller", "VP Finance"],
  q_keywords: "EisnerAmper",
  per_page: 25
)

Parse the response: Each result contains: name, title, company, LinkedIn URL, location, email, and other profile fields. Extract and collect all results into a working list.

Step 3: Evaluate Results

Check how many results from Step 2 match the target titles at the target company.

Quality checks:

1. Filter out results where the company name does not match (fuzzy match is fine -- "EisnerAmper LLP" matches "EisnerAmper")
2. Filter out results where the title does not reasonably match any target title
3. Count remaining high-quality matches

Decision:

• If 3+ quality matches found: skip to Step 7 (Output)
• If fewer than 3 quality matches: proceed to Step 4

Step 4: Fiber Search (Fallback 1 — $0.02/record)

Fiber supports natural-language queries and may have profiles Apollo does not.

Call:

fiber_person_search(
  query: "[title1] OR [title2] OR [title3] at [company_name]",
  page_size: 25
)

After results return:

1. Parse results (extract name, title, company, LinkedIn URL, location)
2. Merge with all previous results
3. Deduplicate by LinkedIn URL

Decision:

• If 3+ total unique quality matches: skip to Step 7 (Output)
• If still fewer than 3: proceed to Step 5

Step 5: Crustdata Structured Search (Fallback 2 — $0.66/page)

Use Crustdata's structured filter search for more precise matching. Run one search per target title, then merge results.

For each target title, call:

crustdata_person_search(
  conditions: [
    {"column": "current_employers.name", "type": "in", "value": "[company_name]"},
    {"column": "current_employers.title", "type": "(.)", "value": "[target_title]"}
  ],
  filter_op: "and",
  limit: 25
)

Example for "Partner" at EisnerAmper:

crustdata_person_search(
  conditions: [
    {"column": "current_employers.name", "type": "=", "value": "EisnerAmper"},
    {"column": "current_employers.title", "type": "(.)", "value": "Partner"}
  ],
  filter_op: "and",
  limit: 25
)

Optional seniority filter: If the user requests senior decision-makers broadly (rather than specific titles), add:

{"column": "current_employers.seniority_level", "type": "in", "value": "VP,C-Level,Director"}

TIP: Use preview: true first to check result count for free before fetching full data.

After all title searches complete:

1. Merge all results into one list
2. Deduplicate by LinkedIn URL (keep the first occurrence)
3. Combine with results from previous steps

Decision:

• If 3+ total unique quality matches: skip to Step 7 (Output)
• If still fewer than 3: proceed to Step 6

Step 6: PDL Search (Fallback 3 — $0.30/record, most expensive)

PeopleDataLabs is the most expensive search provider. Only use as a last resort when other sources have insufficient results.

Call:

pdl_person_search(
  job_titles: ["Partner", "Controller", "VP Finance"],
  company_names: ["EisnerAmper"],
  num_results: 10
)

After results return:

1. Parse results (extract name, title, company, LinkedIn URL, location)
2. Merge with all previous results
3. Deduplicate by LinkedIn URL

Step 7: Output

Present the final deduplicated contact list.

Table format (for the user):

JSON format (for downstream skills):

{
  "company": "EisnerAmper",
  "search_titles": ["Partner", "Controller", "VP Finance"],
  "contacts": [
    {
      "name": "Jane Smith",
      "title": "Partner",
      "company": "EisnerAmper",
      "linkedin_url": "https://linkedin.com/in/janesmith",
      "location": "New York, NY"
    }
  ],
  "total_found": 10,
  "sources": ["apollo", "fiber", "crustdata", "pdl"]
}

Summary line:

Found X contacts matching [titles] at [company]. Sources used: [list of sources that returned results].

If fewer than 3 contacts were found after all fallbacks, tell the user:

Only found X contacts. The company may be small, the titles may be uncommon, or the databases may have limited coverage for this company. Consider broadening the target titles or trying alternate company name spellings.

Gooseworks MCP Tools Reference

Cost Comparison (25 results)

Always start with Apollo. Escalate through Fiber → Crustdata → PDL only as fallbacks.

Tool Details

Crustdata Filter Columns

Examples

Basic: Find Partners and Controllers at EisnerAmper

Find Partners and Controllers at EisnerAmper

Agent calls apollo_person_search with person_titles: ["Partner", "Controller"], q_keywords: "EisnerAmper", per_page: 25.

With more titles: Find VP Finance and CFO at Sage Intacct users

Find VP Finance and CFO at companies using Sage Intacct

Agent builds query: "VP Finance OR CFO at Sage Intacct".

Senior leaders at a specific firm

Find Managing Directors at CPA firms in San Francisco

Agent builds query: "Managing Director at CPA firm San Francisco".

With a LinkedIn URL for disambiguation

Find Partners at EisnerAmper (https://linkedin.com/company/eisneramper)

Agent uses the company name "EisnerAmper" and can use the LinkedIn URL for enrichment if needed.

Troubleshooting

MCP tools not available / connection errors

The Gooseworks MCP tools require the Gooseworks MCP server to be configured in your environment. If you get errors like "tool not found" or connection failures:

1. Check MCP server configuration: Ensure the Gooseworks MCP server is listed in your MCP configuration (e.g., claude_desktop_config.json or equivalent).
2. Server URL: The Gooseworks server must be running and accessible. Check with your workspace admin for the correct server endpoint.
3. Authentication: Gooseworks may require an API key or auth token. Ensure credentials are configured in your MCP server settings.

No results returned

• Try alternate spellings of the company name (e.g., "EisnerAmper" vs "Eisner Amper" vs "EisnerAmper LLP")
• Broaden target titles (e.g., add "Managing Director" alongside "Partner")
• Use the structured search (Step 4) with fuzzy title matching (.) operator
• Try Fiber, Crustdata, or PDL as fallback databases (Steps 4-6)

Too many irrelevant results

• Add more specific title terms rather than broad ones
• Use the structured search with the in operator for exact title matching instead of fuzzy (.)
• Filter results by seniority_level to restrict to senior roles

Duplicate contacts across sources

The skill deduplicates by LinkedIn URL automatically. If you see near-duplicates with slightly different URLs (e.g., trailing slashes), normalize URLs before deduplication by stripping trailing slashes and query parameters.

Metadata

metadata:
  requires:
    mcp_servers: ["gooseworks"]
  cost: "From $0.01 (Apollo) to $7.50 (PDL) depending on provider and result count"