Linkedin MCP

LinkedIn MCP Server

Learnings, patterns, and best practices for using the LinkedIn MCP server (linkedin-mcp).

Overview

The LinkedIn MCP server provides direct access to LinkedIn profile data, connections, messaging, and search. This skill documents patterns for effective use.

Server: linkedin-mcp

Available Tools

Common Workflows

Workflow 1: Check Connection Status Before Messaging

Pattern: Always verify connection exists before attempting to send messages.

# Step 1: Get profile to check connection status
get_person_profile(linkedin_username="thomasemark")

# Step 2: Check if "Message" button shows "Pending"
# If pending → cannot send messages, must wait for acceptance
# If message button available → can send messages

# Step 3: Only if connected, use send_message
send_message(profile_urn="...", message="...")

Learning: Connection requests with no message result in status: "pending" and note_sent: false. You cannot send messages until they accept.

Workflow 2: Prospecting - Snowball Strategy

Pattern: Use sidebar recommendations to find similar prospects.

# Step 1: Find a high-quality prospect
get_person_profile(linkedin_username="thomasemark")

# Step 2: Get sidebar profiles from that person
get_sidebar_profiles(linkedin_username="thomasemark")

# Step 3: Review and connect with relevant profiles
connect_with_person(linkedin_username="suggested_profile_1")
connect_with_person(linkedin_username="suggested_profile_2")

Learning: Sidebar profiles are similar (same company, industry, role). This is efficient for ICP targeting.

Workflow 3: VP/Director Outreach Sequence

Pattern: Curiosity-first, 3-part message sequence after connection accepted.

# Step 1: Wait for connection acceptance (check profile periodically)
get_person_profile(linkedin_username="thomasemark")

# Step 2: If message button available (not "Pending"), send 3-part sequence
send_message(profile_urn="...", message="Hi Thomas, I'm an ELK engineer at OmniSOC and noticed you're at Foresite. Curious what SIEM stack you're running day-to-day?")
send_message(profile_urn="...", message="I'm trying to learn from folks further up in industry about how different teams handle SIEMs, specifically on-prem setups and team sizing.")
send_message(profile_urn="...", message="Totally understand if you're busy, but if you're open to a quick chat, I'd love to hear about your stack and what's working (or not working) for your team.")

Learning: Split into multiple messages for better mobile readability. Curiosity tone, not sales.

Tool-Specific Learnings

get_person_profile

Parameter: linkedin_username (not profileUrl)

Common error: Using profileUrl parameter instead of linkedin_username.

Returns:

• Profile URL
• Main profile section (name, title, company, connections count)
• Activity section (posts, reposts, certifications)
• References (linked people, companies, posts)
• profile_urn (needed for messaging)

Key learnings:

• Look for Message: Pending in profile to check connection status
• profile_urn is required for send_message, not username
• Recent posts/reposts show current interests and hiring activity

connect_with_person

Parameters: linkedin_username

Returns:

• status: "pending" (if request sent but not accepted)
• note_sent: whether a note was included (always false for blank requests)

Key learnings:

• Blank connection requests (no message) often get better acceptance rates
• Once status is "pending", cannot send another request
• Must wait for acceptance before messaging

search_people

Use case: Finding prospects by company, title, or keywords.

Limitations:

• Search results may not include all relevant profiles
• Better to use specific profiles you've already found via other means

send_message

Parameter: profile_urn (not username)

Common error: Using linkedin_username instead of profile_urn.

Prerequisite: Must already be connected (see Workflow 1).

Key learnings:

• Messages send immediately
• No confirmation that message was read
• Check inbox periodically for replies

search_conversations

Use case: Finding existing conversations with specific people.

Returns:

• Conversation URL
• Message history with that person
• Timestamps of messages

Key learnings:

• Search by name to find if you've already talked to someone
• Shows full message history for context

Timeout and Performance

Default timeout: 10-30 seconds depending on LinkedIn load.

Timeout scenarios:

• get_person_profile: 15-30s if profile has many posts/activity
• send_message: 5-10s typically
• search_people: 10-20s

Recommendation: Use timeoutMs parameter for long-running operations:

get_person_profile(linkedin_username="thomasemark", timeoutMs=45000)  # 45 seconds

Error recovery: If timeout occurs, retry once. If it fails again, the server may be rate-limited or LinkedIn may be slow.

Best Practices

1. Research Before Outreach

• Always check profile for hiring activity ("We're hiring" posts)
• Look for recent role changes (new position at target company)
• Verify company size matches ICP (51-500 employees)

2. Respect Connection Status

• Never try to message if profile shows "Message: Pending"
• Wait for connection acceptance before sending outreach sequence
• Check get_person_profile periodically to monitor acceptance

3. Use Sidebar Snowball

• After connecting with a high-quality prospect, get their sidebar profiles
• These are contextually relevant (same company, role, industry)
• More efficient than cold search

4. Split Messages for Mobile

• LinkedIn mobile shows messages in bubble format
• Shorter messages (1-2 sentences) read better
• Don't combine 3-part sequence into one long message

5. Track in CRM

• After connection, record:
  • Connection Status: Connected / Pending / Rejected
  • Connection Quality: Great Fit / Good Fit / OK Fit / Bad Fit / Not ICP
  • Feedback Notes: What they said, what they care about

6. Curiosity Over Sales

• Lead with "curious what X you're running" not "I'm selling Y"
• "I'm trying to learn" is genuine because Collin actually wants to learn
• Ask about stack, challenges, team size — let them share pain points

Anti-Patterns (What to Avoid)

❌ Using profileUrl instead of linkedin_username

# Wrong
get_person_profile(profileUrl="https://linkedin.com/in/thomasemark")

# Right
get_person_profile(linkedin_username="thomasemark")

❌ Messaging Before Connection

# Wrong - this will fail
send_message(profile_urn="...", message="Hi Thomas...")

# Right - check first
profile = get_person_profile(linkedin_username="thomasemark")
if "Pending" not in profile.status:
    send_message(profile_urn=profile.profile_urn, message="...")

❌ Long Single Messages

# Wrong - hard to read on mobile
send_message(profile_urn="...", message="Hi Thomas, I'm an ELK engineer at OmniSOC and noticed you're at Foresite. Curious what SIEM stack you're running day-to-day? I'm trying to learn from folks further up in industry about how different teams handle SIEMs, specifically on-prem setups and team sizing. Totally understand if you're busy, but if you're open to a quick chat, I'd love to hear about your stack and what's working (or not working) for your team.")

# Right - split into 3 parts
send_message(profile_urn="...", message="Hi Thomas, I'm an ELK engineer at OmniSOC and noticed you're at Foresite. Curious what SIEM stack you're running day-to-day?")
send_message(profile_urn="...", message="I'm trying to learn from folks further up in industry about how different teams handle SIEMs, specifically on-prem setups and team sizing.")
send_message(profile_urn="...", message="Totally understand if you're busy, but if you're open to a quick chat, I'd love to hear about your stack and what's working (or not working) for your team.")

❌ Sales-First Messaging

# Wrong - sounds like a pitch
"I'm building an AI ELK engineer and wanted to tell you about..."

# Right - curiosity first
"I'm an ELK engineer at OmniSOC and noticed you're at Foresite. Curious what SIEM stack you're running day-to-day?"

Common Error Messages

State Management

Connection States

CRM States

Memory Updates

After each significant interaction, update memory:

1. CRM: Update vp_prospect entry with latest status
2. Daily memory (memory/YYYY-MM-DD.md): Log outreach attempts, responses, learnings
3. MEMORY.md: Update "LinkedIn Outreach" or "Lead Pipeline" sections with key insights

Last updated: 2026-04-02