MCP Data Enrichment

Leadspicker Data Enrichment (MCP)

Prepares and enriches contact data in Leadspicker projects through the Leadspicker MCP server. Each enrichment creates a magic column that Leadspicker populates asynchronously for every contact in the project.

This skill works in two modes:

1. Selective mode — user names specific columns (e.g., "enrich LinkedIn company description and website summary"). Run only those, plus any prerequisites they depend on.
2. Pipeline mode — user describes a goal (e.g., "prepare data for outreach"). Pick a recommended pipeline.

MCP tools used

If you need an endpoint not covered above, use mcp__leadspicker__leadspicker_api_call and introspect via mcp__leadspicker__leadspicker_api_spec. For everything in this skill the curated tools are enough.

User input

Always confirm these before running anything:

Project resolution

Display format: always render projects as {name} (#{id}) — name first, ID in parentheses.

1. User named a project → call mcp__leadspicker__list_projects(search_query="{name}", limit=5). If multiple match, ask which one.
2. User did not name a project → call mcp__leadspicker__list_projects(order_by="last_active", limit=20) and ask the user to pick.

Use the resolved project's id for subsequent calls.

⚠️ Credit discipline

Every enrichment column costs credits. Default to a lean minimum, never auto-enrich "just in case". Add optional columns only when the user's persona / criteria explicitly require them.

Lean default for outreach prep:

Hard rules:

• NEVER default to li_company_employee_count. Always prefer li_company_size (range like 51-200).
• NEVER default to li_latest_posts. It is the most expensive enrichment.
• Before launching, parse the user's request for triggers (geography, size, department, age) and include the matching optional column only if triggered.

Dependency map

Some columns need other columns to exist first. When the user requests a column, auto-include its prerequisites.

No prerequisites:
  company_name, company_website, company_linkedin,
  person_linkedin, li_full_name, person_position, person_country,
  li_location, li_city, li_company_size, li_company_country,
  company_founded_year, li_company_employee_count,
  enrich_emails, linkedin_premium, is_first_degree_connection,
  followers_count, linkedin_skills

Needs person_linkedin:
  li_about_me, li_latest_posts, present_experience, past_experience,
  education_summary

Needs company_linkedin:
  li_company_description, all li_company_people_* department headcounts

Needs company_website:
  website_text_summary

Needs enrich_emails:
  validate_emails, email_mx

Needs data from earlier enrichments (for AI context):
  any ai_custom_column (classification, icebreakers, cleaning, …)

Selective resolution example. User: "I want LinkedIn Company Description and Company Website Summary":

1. li_company_description needs company_linkedin
2. website_text_summary needs company_website
3. Neither prerequisite has its own prerequisite

Result — run in this order:

• Batch 1 (prerequisites): company_linkedin, company_website
• Batch 2 (requested): li_company_description, website_text_summary

Always show this resolved plan to the user before executing.

Available enrichment columns

Each enrichment is created with mcp__leadspicker__enrich_contacts(project_id, magic_column_type, column_name). AI columns also pass prompt and is_boolean. Built-in columns don't need a prompt.

Person data (from LinkedIn)

Company data

Department headcounts

Email

AI-powered columns (custom)

magic_column_type: "ai_custom_column" with a prompt. They reference any variable from the columns above. For prompt-writing details, use the mcp-classification skill.

Common AI enrichments: Clean First Name, Clean Company Name, Job Title Cleaning, Company Industry, B2B/B2C, Icebreaker — Latest LI Posts, LinkedIn Post Summarization, Targeted Job Titles, First Name Vocative — CZ, Last Name Vocative — CZ.

Other built-in types

Recommended pipelines

Order matters — later steps often depend on earlier ones. Show the picked pipeline to the user, confirm, then execute.

Lean Outreach Prep — DEFAULT for "prepare for outreach"

Minimum column set for classification + personalization + email outreach. Start here; add optional columns from the credit-discipline table only when triggered.

Batch 1 — Core Identity (ALWAYS):
  company_name, company_website, company_linkedin,
  person_linkedin, li_full_name, person_position

Batch 2 — Classification context (ALWAYS):
  li_company_description, website_text_summary

Batch 2b — Conditional company-level (only if triggered):
  li_company_country     — user cares where the company is based
  li_company_size (RANGE) — user cares about employee category / size band
  li_company_people_*    — user mentioned that department
  company_founded_year   — user mentioned company age / stage
  li_company_employee_count — only if user explicitly asked for exact count

Batch 3 — Conditional person-level (only if triggered):
  person_country         — user cares where the person is based
  li_about_me            — personalization depends on bio
  present_experience     — personalization uses role tenure
  li_latest_posts        — icebreaker uses post activity (EXPENSIVE — confirm first)

Batch 4 — Email (ALWAYS for email outreach):
  enrich_emails, then validate_emails

Batch 5 — AI columns:
  any classification / scoring / personalization (ai_custom_column).
  Hand off to mcp-classification or mcp-personalizer skills.

Show the plan with three sections: Always, Added because you mentioned X, NOT enriching (saves credits). Only proceed after explicit confirmation.

Quick email enrichment

Batch 1: enrich_emails
Batch 2: validate_emails (after batch 1 completes)

Company research

Batch 1: company_name, company_website, company_linkedin
Batch 2: li_company_description, website_text_summary,
         li_company_size, company_founded_year (li_company_employee_count only if asked)
Batch 3: li_company_people_* as needed

LinkedIn person enrichment

Batch 1: person_linkedin, li_full_name, person_position
Batch 2: li_about_me, li_latest_posts, present_experience,
         past_experience, education_summary, linkedin_skills

Outreach prep (CZ market)

Batch 1–4: Lean Outreach Prep (with credit discipline)
Batch 5:   AI custom columns:
  Clean First Name, First Name Vocative — CZ, Clean Company Name,
  any classification columns,
  Icebreaker — Latest LI Posts (only if li_latest_posts was enriched)

For prompt details on AI columns, use mcp-personalizer (vocatives, hooks, salutations) or mcp-classification (boolean / string / scored classifications).

Full historical enrichment (legacy)

Use only when the user explicitly says "run full enrichment" and reconfirms after the cost warning. Otherwise prefer Lean Outreach Prep.

All of Lean Outreach Prep, plus:
  li_company_size, li_company_country, company_founded_year,
  li_company_employee_count, li_about_me, li_latest_posts,
  present_experience, li_city, person_country

Workflow

Step 1 — collect input

Resolve the project (see Project resolution) and gather the enrichment request.

Step 2 — determine mode and build the plan

Selective mode (user named specific columns):

1. Match the user's words to columns in the tables above
2. Look up the magic_column_type (this is what mcp__leadspicker__enrich_contacts needs)
3. Check the dependency map — add missing prerequisites
4. Group into batches: prerequisites first, then requested
5. Present the plan separating prerequisites from requested columns

Example to show user:

I'll enrich these columns for Marketing Q4 (#26256):

Prerequisites (needed first):
  • Company LinkedIn Profile (company_linkedin)
  • Company Website (company_website)

Your requested columns:
  • LinkedIn Company Description (li_company_description)
  • Company Website Summary (website_text_summary)

Total: 4 enrichments in 2 batches. OK to proceed?

Pipeline mode (user described a goal):

1. Pick the best matching recommended pipeline
2. Apply credit-discipline triggers (geography, size, department, age, posts)
3. Present the full pipeline with all batches
4. Show what's NOT enriched and why (credit-saving)

Step 3 — confirm with user

Before any enrichment call, show the plan and get explicit approval. For AI columns, show the prompt text or hand off to mcp-classification / mcp-personalizer.

Step 4 — execute

For each batch, call mcp__leadspicker__enrich_contacts once per column. Wait briefly between calls (≈1 s) to avoid rate-limits.

mcp__leadspicker__enrich_contacts(
    project_id=26256,
    magic_column_type="company_website",
    column_name="Company Website",
)

For an AI custom column:

mcp__leadspicker__enrich_contacts(
    project_id=26256,
    magic_column_type="ai_custom_column",
    column_name="Company Industry",
    prompt="Determine the primary industry of this company.\n\n…\n\nInput:\n- Company Name: {{company_name}}\n- Company Website: {{company_website}}\n- LinkedIn Company Description: {{linkedin_company_description}}\n- Company Website Summary: {{website_text_summary}}",
    is_boolean=False,
)

Escape newlines in prompt as \n in the JSON payload that the MCP tool builds. Template variables like {{company_name}} are filled by Leadspicker server-side.

For pipelines where later batches genuinely need earlier data (e.g. validate_emails after enrich_emails, AI columns after enrichment), ask the user to confirm the previous batch completed in the Leadspicker UI before launching the next one.

Step 5 — report summary

After the batch finishes, report:

• Columns created (mark prerequisites vs requested)
• Any failures with reason
• Suggested next steps:
  • After company data → "You can now run AI classification (mcp-classification)."
  • After email enrichment → "Validate emails next."
  • After LinkedIn person data → "Good base for icebreaker / personalization (mcp-personalizer)."

Natural language → column matching

Users won't always use exact column names. Map common phrases:

When in doubt, show the column list and ask the user to pick.

Common errors

Safety

• Always show the plan to the user for approval before launching a batch.
• For AI columns, always show the prompt before launching.
• Default to lean enrichment; only add optional columns when explicitly triggered by user input.