MCP Personalizer

Leadspicker Personalizer (MCP)

Generates short personalization fragments for cold email campaigns in Leadspicker projects through the Leadspicker MCP server. Each column runs GPT-4o-Mini on every contact and produces a short sentence fragment or a grammar-correct salutation.

This skill is a companion to mcp-data-enrichment (provides input data) and mcp-classification (filters contacts).

Core philosophy

Leadspicker never generates full messages — only short parts.

Every personalization output is a short sentence fragment (typically 14–16 words). This keeps GPT-4o-Mini focused, reduces hallucination, and gives the user full control over the final message structure. The exact word limit can be adjusted per request.

Examples:

• "I saw your company specializes in AI-powered sales automation" (website hook)
• "I noticed your recent post about hiring challenges in tech" (icebreaker)
• "Dobrý den, pane Nováku" (Czech salutation)
• "Tomáši" (Czech first-name vocative)

These fragments are inserted into message templates alongside static text the user controls.

MCP tools used

User input

Project resolution

Display format: always render projects as {name} (#{id}).

1. User named a project → mcp__leadspicker__list_projects(search_query="{name}", limit=5). Confirm; if multiple, ask which one.
2. No named project → mcp__leadspicker__list_projects(order_by="last_active", limit=20) and ask the user to pick.

If the user did not provide column_name, generate one from the type + language (e.g. "Website Hook", "CZ Salutation", "Icebreaker - DE").

System prompt — fixed, do not modify

All personalization columns use string mode (is_boolean=False).

Server forces: {"result": "response"}

Implications:

• NEVER include output-format instructions — the system prompt enforces JSON.
• Just describe the task, rules, and input variables.
• Output is wrapped in {"result": "..."} automatically.

Personalization types

A. Company-based hooks

Short sentences referencing what the company does.

Website hook

References what the company does based on website content. Always include {{linkedin_company_description}} as a fallback — when website data is empty, LinkedIn fills the gap.

• Variables: {{company_name}}, {{website_text_summary}}, {{linkedin_company_description}}
• Typical output: "I saw your company helps retailers optimize their supply chain"

LinkedIn company hook

References the company's LinkedIn description (mission / positioning).

• Variables: {{company_name}}, {{linkedin_company_description}}
• Typical output: "I noticed your team is building AI tools for sales teams"

Combined company hook

Uses both website and LinkedIn for richest context.

• Variables: {{company_name}}, {{company_website}}, {{linkedin_company_description}}, {{website_text_summary}}

B. Person-based hooks

Experience hook

References the person's role / background / career trajectory.

• Variables: {{full_name}}, {{position}}, {{linkedin_about_me}}, {{present_experiences}}
• Typical output: "With your experience leading sales teams at enterprise companies"

LinkedIn posts icebreaker

References something the person recently posted on LinkedIn.

• Variables: {{linkedin_latest_posts}}
• Typical output: "I noticed your post about the challenges of scaling outbound"

Education / skills hook

References education or specific expertise.

• Variables: {{full_name}}, {{education_summary}}, {{linkedin_skills}}
• Typical output: "With your background in data science and machine learning"

C. Language-specific columns

Salutation — formal greeting with declined name

• Variables: {{full_name}}, {{position}}
• Czech: "Dobrý den, pane Nováku" / "Dobrý den, paní Nováková"
• German: "Sehr geehrter Herr Schmidt" / "Sehr geehrte Frau Müller"
• Polish: "Szanowny Panie Kowalski" / "Szanowna Pani Kowalska"

First-name vocative — declined first name

• Variables: {{first_name}}
• Czech: "Tomáši" from Tomáš, "Kateřino" from Kateřina
• Polish: "Tomaszu" from Tomasz, "Katarzyno" from Katarzyna

Last-name vocative — declined last name with title

• Variables: {{last_name}}, {{full_name}}, {{position}}
• Czech: "pane Blaťáku" / "paní Blaťáková"

Smart variable selection

Prompt-writing rules

1. Word limit always. Every prompt must contain: "Keep your output under 16 words." Single most important rule — controls hallucination.
2. Natural tone. Should sound like something a human would write. Never salesy.
3. Sentence fragment, not full sentence. Hooks are part of a larger message.
4. Fallback for missing data. "If there is not enough information, return empty string."
5. No output-format instructions. System prompt handles JSON.
6. One task per prompt. Don't combine hook + salutation in one column.
7. Language instruction. For non-English output, append at the end: "Write your output in {language}."

Hook-specific rules

• Reference something specific from the input — never generic.
  • Bad: "I see your company is doing great things."
  • Good: "I saw your company helps SaaS teams reduce churn with AI."
• Start with natural openers: "I saw...", "I noticed...", "With your...", "Given your...".
• Don't add a transition phrase — the user controls what follows.
• Focus on business-relevant content; skip vacations / birthdays.

Vocative / salutation rules

• Include explicit declension rules with examples in the prompt.
• Include gender-detection logic (name endings, position keywords).
• Czech: masculine names typically end in consonant, feminine in -a / -á / -e.
• Provide fallback for ambiguous names — use the more formal / neutral option.
• Never use diminutive forms — always formal / standard vocative.

Prompt templates

Company website hook

column_name: Website Hook (or Website Hook - {LANG})  ·  is_boolean=False

⚠️ Always include {{linkedin_company_description}} alongside {{website_text_summary}}.

Write a short personalization sentence for a cold email that references what this company does.

Rules:
- Keep your output under 16 words
- Start with "I saw" or "I noticed" followed by a specific reference to the company's activity
- Use the most interesting and specific detail from the available data
- Reference a product, service, market, technology, or initiative — be concrete
- Do NOT be generic (no "I see your company is doing great things")
- Do NOT include a call to action or transition phrase
- The tone should be natural and conversational, not salesy
- If there is not enough information, return empty string

Input:
- Company Name: {{company_name}}
- Company Website Summary: {{website_text_summary}}
- LinkedIn Company Description: {{linkedin_company_description}}

For non-English output, append at the end:

Write your output in {language}.

Native-language opener equivalents:

• German: "Ich habe gesehen" / "Mir ist aufgefallen"
• Polish: "Zauważyłem" / "Zauważyłam"
• Czech: "Všiml jsem si" / "Zaujalo mě"

LinkedIn company hook

column_name: Company Hook  ·  is_boolean=False

Write a short personalization sentence for a cold email that references this company's focus or mission based on their LinkedIn description.

Rules:
- Keep your output under 16 words
- Start with "I noticed" or "I saw" followed by a specific reference to the company
- Reference their mission, product focus, or market positioning
- Do NOT be generic or vague
- Do NOT include a call to action or transition phrase
- The tone should be natural and conversational, not salesy
- If there is not enough information, return empty string

Input:
- Company Name: {{company_name}}
- LinkedIn Company Description: {{linkedin_company_description}}

Person experience hook

column_name: Experience Hook  ·  is_boolean=False

Write a short personalization sentence for a cold email that references this person's professional background or current role.

Rules:
- Keep your output under 16 words
- Start with "With your experience" or "Given your background" or similar natural opener
- Reference something specific from their role, expertise, or career focus
- Do NOT mention the person's name in the output
- Do NOT be generic
- Do NOT include a call to action or transition phrase
- Focus on business-relevant aspects of their experience
- If there is not enough information, return empty string

Input:
- Position: {{position}}
- LinkedIn About Me: {{linkedin_about_me}}
- Present Positions Summary: {{present_experiences}}

LinkedIn posts icebreaker

column_name: Icebreaker  ·  is_boolean=False

Write a personalized icebreaker sentence for a cold email based on the person's latest LinkedIn post.

Rules:
- Keep your output under 16 words
- Start with "I noticed" or "I saw"
- If the input is a personal post, use "I noticed on your LinkedIn..."
- If the input starts with "Company post:", use "I noticed on your company LinkedIn..."
- Focus on business-related content (product launches, insights, industry opinions)
- Skip personal posts about vacations, birthdays, hobbies — if only personal content exists, return empty string
- The tone should be natural and not salesy
- Do NOT include a call to action or transition phrase
- If the input is empty or has no meaningful business content, return empty string

Input:
- LinkedIn Latest Posts: {{linkedin_latest_posts}}

Education / skills hook

column_name: Skills Hook  ·  is_boolean=False

Write a short personalization sentence for a cold email that references this person's education or specific expertise.

Rules:
- Keep your output under 16 words
- Start with "With your background in" or "Given your expertise in" or similar natural opener
- Reference a specific field of study, skill, or area of expertise
- Do NOT mention the person's name in the output
- Do NOT be generic
- Do NOT include a call to action or transition phrase
- If there is not enough information, return empty string

Input:
- Education Summary: {{education_summary}}
- LinkedIn Skills: {{linkedin_skills}}

Language-specific templates

Czech salutation

column_name: CZ Salutation  ·  is_boolean=False

Generate a formal Czech salutation for a business email.

Your task:
1. Determine the person's gender from their name and job title
2. Create the correct formal greeting in Czech

Gender detection rules:
- Feminine indicators: first name ends in -a, -e, or -ie (Jana, Marie, Kateřina, Lucie); last name ends in -ová, -á (Nováková, Černá)
- Masculine indicators: first name ends in a consonant or -o (Jan, Tomáš, Michal, Marek); last name without -ová/-á suffix
- Position can help: "ředitelka", "manažerka", "náměstkyně" indicate female; "ředitel", "manažer", "náměstek" indicate male
- If ambiguous, default to masculine form

Czech salutation rules:
- Masculine: "Dobrý den, pane" + last name in vocative case
  - -ák → -áku (Novák → Nováku, Dvořák → Dvořáku)
  - -ek → -ku (Voříšek → Voříšku, Havlíček → Havlíčku)
  - -ec → -če (Němec → Němče)
  - -ský/-cký → unchanged (Novotný → Novotný)
  - -ka → -ko (Procházka → Procházko, Růžička → Růžičko)
  - -da → -do (Svoboda → Svobodo)
  - general consonant → add -e (Vlach → Vlache)
- Feminine: "Dobrý den, paní" + last name unchanged (Nováková → Nováková)

Return only the salutation, nothing else.
Examples: "Dobrý den, pane Nováku", "Dobrý den, paní Dvořáková", "Dobrý den, pane Blaťáku"

If the name is not recognizable or empty, return empty string.

Input:
- Full Name: {{full_name}}
- Position: {{position}}

Czech first-name vocative

column_name: CZ First Name Vocative  ·  is_boolean=False

You are an expert in Czech grammar, specifically declension rules for the vocative case (5. pád).

Given a Czech first name:
1. Add any missing diacritics (háčky and čárky)
2. Transform the name into the correct vocative form

Masculine vocative rules:
- -áš → -áši (Tomáš → Tomáši, Lukáš → Lukáši)
- -an → -ane (Jan → Jane, Štěpán → Štěpáne)
- -el → -ele (Michal → Michale, Pavel → Pavle)
- -ek → -ku (Marek → Marku, Zdeněk → Zdeňku)
- -in → -ine (Martin → Martine)
- -ír → -íře (Vladimír → Vladimíre)
- -ej → -eji (Ondřej → Ondřeji, Matěj → Matěji)
- -tr → -tře (Petr → Petře)
- -ip → -ipe (Filip → Filipe)
- -ub → -ube (Jakub → Jakube)
- -id → -ide (David → Davide)
- -os → -oši (Miloš → Miloši)
- -áv → -áve (Václav → Václave)
- -o ending: stays the same (Ivo → Ivo)

Feminine vocative rules:
- -a → -o (Jana → Jano, Petra → Petro, Simona → Simono, Kateřina → Kateřino)
- -ie/-e → stays the same (Marie → Marie, Lucie → Lucie)

IMPORTANT: Use only the standard (official) vocative form, never diminutives.
- "Marie" → "Marie" (NOT "Maruško")
- "Simona" → "Simono" (NOT "Simonko")
- "Kateřina" → "Kateřino" (NOT "Katko")

Return only the vocative form, nothing else. If no name is provided, return empty string.

Input:
- First Name: {{first_name}}

Czech last-name vocative

column_name: CZ Last Name Vocative  ·  is_boolean=False

You are an expert in Czech grammar. Generate the vocative form (5. pád) of a Czech last name with the appropriate title (pane/paní).

Your task:
1. Determine the person's gender from their full name and position
2. Apply the correct vocative declension to the last name
3. Return the title + declined last name

Gender detection:
- Feminine: first name ends in -a/-e/-ie (Jana, Marie, Lucie); last name ends in -ová/-á (Nováková)
- Masculine: first name ends in consonant/-o (Jan, Tomáš, Michal); last name without -ová
- Position keywords: -ka/-kyně suffix = female (manažerka, ředitelka), no suffix = male (manažer, ředitel)

Masculine last-name vocative rules:
- -ák → -áku (Novák → Nováku, Dvořák → Dvořáku, Blaťák → Blaťáku)
- -ek → -ku (Havlíček → Havlíčku, Voříšek → Voříšku)
- -ec → -če (Němec → Němče)
- -ík → -íku (Sedlačík → Sedlačíku)
- -ka → -ko (Procházka → Procházko, Růžička → Růžičko)
- -da → -do (Svoboda → Svobodo)
- -ý/-í → unchanged (Novotný → Novotný, Malý → Malý)
- general consonant → add -e (Vlach → Vlache)
- Output format: "pane {vocative_last_name}"

Feminine last-name rules:
- Keep the last name unchanged (nominative = vocative)
- Output format: "paní {last_name}"

Return only the title + last name (e.g., "pane Nováku" or "paní Dvořáková"), nothing else.
If the name is not recognizable or empty, return empty string.

Input:
- Last Name: {{last_name}}
- Full Name: {{full_name}}
- Position: {{position}}

Polish salutation

column_name: PL Salutation  ·  is_boolean=False

Generate a formal Polish salutation for a business email.

Your task:
1. Determine the person's gender from their first name and last name
2. Create the correct formal greeting in Polish

Gender detection rules:
- Feminine indicators: first name ends in -a (Anna, Katarzyna, Monika, Agnieszka, Magdalena, Joanna, Ewa, Dorota, Aleksandra, Natalia); last name ends in -ska, -cka, -dzka, -wska (Kowalska, Nowacka)
- Masculine indicators: first name ends in a consonant or -o (Jan, Tomasz, Michał, Marek, Piotr, Krzysztof, Andrzej, Adam); last name ends in -ski, -cki, -dzki, -wski or consonant (Kowalski, Nowak)
- If ambiguous, default to masculine form

Polish salutation format:
- Male: "Szanowny Panie" + last name in nominative (unchanged)
  - Kowalski → "Szanowny Panie Kowalski"
- Female: "Szanowna Pani" + last name in nominative (unchanged)
  - Kowalska → "Szanowna Pani Kowalska"

Extract the last name from the full name (last word in the name).

Return only the salutation, nothing else.
Examples: "Szanowny Panie Kowalski", "Szanowna Pani Nowak"

If the name is not recognizable or empty, return empty string.

Input:
- Full Name: {{full_name}}
- Position: {{position}}

German salutation

column_name: DE Salutation  ·  is_boolean=False

Generate a formal German salutation for a business email.

Your task:
1. Determine the person's gender from their first name
2. Create the correct formal greeting in German

Gender detection rules:
- Common male first names: Hans, Peter, Thomas, Michael, Stefan, Andreas, Martin, Marco, Alexander, David, Daniel, Christian, Markus, Frank, Klaus, Jürgen, Wolfgang, Bernd, Ralf, Uwe, Dirk, Oliver, Patrick, Sven, Tobias, Florian, Matthias, Christoph
- Common female first names: Anna, Maria, Marie, Claudia, Sabine, Petra, Monika, Susanne, Andrea, Karin, Julia, Sandra, Nicole, Katharina, Stefanie, Birgit, Heike, Martina, Simone, Christina, Laura, Lisa, Sarah, Elena, Nadine, Anja, Tanja, Silke, Melanie
- Names ending in -a, -e, -ie, -ine are typically female
- Names ending in consonant are typically male
- If ambiguous, use "Sehr geehrte/r" as neutral form

German salutation format:
- Male: "Sehr geehrter Herr {last_name}"
- Female: "Sehr geehrte Frau {last_name}"
- Last name stays unchanged (no declension in German)

Extract the last name from the full name (last word in the name).

Return only the salutation, nothing else.
Examples: "Sehr geehrter Herr Schmidt", "Sehr geehrte Frau Müller"

If the name is not recognizable or empty, return empty string.

Input:
- Full Name: {{full_name}}
- Position: {{position}}

Generic salutation template (any language)

When the user requests a salutation in a language not covered above:

Generate a formal {language} salutation for a business email.

Your task:
1. Determine the person's gender from their name and job title
2. Create the correct formal greeting in {language}

{LANGUAGE_SPECIFIC_GENDER_RULES}
{LANGUAGE_SPECIFIC_SALUTATION_FORMAT}

Return only the salutation, nothing else.
If the name is not recognizable or empty, return empty string.

Input:
- Full Name: {{full_name}}
- Position: {{position}}

Sample formats:

• French: "Cher Monsieur {last_name}" / "Chère Madame {last_name}" (no declension)
• Spanish: "Estimado Sr. {last_name}" / "Estimada Sra. {last_name}" (no declension)

Multi-language system

Language selection

1. User specifies the language in their request: "website hook in German", "Czech salutation", "icebreaker in Polish".
2. Default is English when no language specified.
3. For hooks: append "Write your output in {language}." at the end of the prompt.
4. For vocatives / salutations: use the language-specific template with built-in grammar rules.

Capabilities

Auto-generate salutation with hooks

Whenever a user requests any hook in a specific language, automatically also prepare the salutation column for that language. A salutation is fundamental to outreach — the user should never have to ask for it separately.

For languages with vocatives (Czech, Polish, Slovak) also auto-prepare:

• {LANG} Salutation
• {LANG} First Name Vocative
• {LANG} Last Name Vocative

For languages without vocatives (German, French, Spanish, English) auto-prepare:

• {LANG} Salutation (Herr/Frau, Monsieur/Madame, Sr./Sra.)

The user can opt out by saying "just the hook" or "no salutation".

Workflow

Step 1 — collect input

Resolve the project (see Project resolution) and gather the personalization request and target language. If column_name is missing, generate one from type + language.

Step 2 — determine type

1. Type: company hook / person hook / icebreaker / salutation / vocative
2. Variables: pick from Smart variable selection
3. Language: default English, or user-specified
4. Auto-include salutation for any non-English hook request (see Auto-generate above)

Step 3 — write the prompt

Follow the rules above:

1. Clear task definition
2. Word-limit rule (≤ 16 words for hooks)
3. Tone / style / content rules
4. Fallback for missing data
5. Input variables (only relevant ones)
6. Language instruction if non-English

Step 4 — show prompt to user

Show:

• column_name
• Full prompt text
• Target language

Ask for approval.

Step 5 — preview on first 5 rows

mcp__leadspicker__preview_ai_prompt(
    project_id=29842,
    column_name="Website Hook - CZ",
    prompt="Write a short personalization sentence …\n…\nInput:\n- Company Name: {{company_name}}\n…\n\nWrite your output in Czech.",
    is_boolean=False,
    magic_column_type="ai_custom_column",
)

While that processes, fetch contact context:

mcp__leadspicker__list_contacts(project_id=29842, page=1, page_size=5)

Show a combined table:

Preview (first 5 rows):

| Name           | Position       | Company        | Result                                           |
|----------------|----------------|----------------|--------------------------------------------------|
| Tomáš Blaťák   | Sales Manager  | Leadspicker    | Všiml jsem si, že vaše firma pomáhá …            |
| Jana Nováková  | CEO            | Acme s.r.o.    | Všimla jsem si, že vaše firma se specializuje … |
…

Does this look correct? Should I launch for all contacts?

If changes are needed → adjust prompt and re-preview. If cancel → stop.

Step 6 — launch full personalization

mcp__leadspicker__enrich_contacts(
    project_id=29842,
    magic_column_type="ai_custom_column",
    column_name="Website Hook - CZ",
    prompt="<same prompt>",
    is_boolean=False,
)

Step 7 — report and suggest next steps

• Confirm column was created and is processing.
• Suggest related personalizations:
  • After website hook → "Want a LinkedIn company hook too?"
  • After icebreaker → "Want a hook based on their experience?"
  • After salutation → "Want first-name and last-name vocatives too?"
  • After CZ columns → "Want hooks in Czech as well?"

Batch mode — personalization packages

When the user asks for "full personalization" or "all hooks":

English package

Column 1: Website Hook
Column 2: Company Hook (LinkedIn)
Column 3: Experience Hook
Column 4: Icebreaker

Czech package

Column 1: Website Hook - CZ
Column 2: Company Hook - CZ
Column 3: Experience Hook - CZ
Column 4: Icebreaker - CZ
Column 5: CZ Salutation
Column 6: CZ First Name Vocative
Column 7: CZ Last Name Vocative

Any-language package

Column 1: Website Hook - {LANG}
Column 2: Company Hook - {LANG}
Column 3: Experience Hook - {LANG}
Column 4: Icebreaker - {LANG}
Column 5: {LANG} Salutation (always)
Column 6: {LANG} First Name Vocative (vocative languages only)
Column 7: {LANG} Last Name Vocative (vocative languages only)

Execute via mcp__leadspicker__enrich_contacts sequentially with ≈1 s delay between calls. Preview the first column to verify quality, then launch all after approval.

Natural language → personalization matching

Common errors

Safety

• Always show the prompt to the user before previewing or launching.
• Always preview before launching.
• Write all prompts in English for best GPT-4o-Mini performance, with "Write your output in {language}." appended for non-English output. The exception is vocative output (Czech / Polish / Slovak grammar rules ARE in the prompt itself).