MCP Robots

Leadspicker Robots (MCP)

Creates and launches robots (LinkedIn scraping automations) in Leadspicker through the Leadspicker MCP server via mcp__leadspicker__create_robot. Each robot scrapes LinkedIn data and deposits the results as leads into a target Leadspicker project.

4 supported robot types:

MCP tools used

User input

If anything is missing, ask for ALL missing parameters in a single follow-up — not one by one.

Project resolution

Display format: always render 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:

   Which project should receive the scraped leads?
   1. CEOs Czechia (#30877) — last active 2026-03-23
   2. SaaS Founders (#30512) — last active 2026-03-21
   …
   New: create a new project
   

3. User wants a new project → ask for a name, then mcp__leadspicker__create_project(name="{name}", timezone="Europe/Prague"). Use the returned id.

URL detection — auto-routing to robot type

Rules:

• Profile URLs (/in/... or /company/...) NOT inside a search URL → li_posts_monitor.
• If ambiguous, ask the user which type they intend.
• Multiple URLs of the same type are fine — concatenate them with newlines in robot_urls.
• Do NOT mix URL types in a single robot — for mixed input, create separate robots.

Sales Navigator URL — how to obtain

This MCP skill does not generate Sales Navigator URLs from filter descriptions. If the user needs a Sales Nav URL:

1. Ask them to build it manually in LinkedIn Sales Navigator (set filters, copy the URL).
2. Or ask them to paste an existing URL.
3. Or, when running outside MCP, generate it programmatically with their preferred tool.

Then use the URL as robot_urls input here.

Robot type details

1) li_posts_extract — scrape specific posts

Scrapes likers and commenters from specific LinkedIn post URLs.

URL input: one or more post URLs (/feed/update/ or /posts/).

Questions to ask the user: none beyond the URLs. Use defaults and proceed.

2) li_posts_monitor — track posts by account

Monitors posts published by specific LinkedIn profiles or company pages.

URL input: one or more profile URLs (/in/...) or company URLs (/company/...).

Ask the user: how far back to look (max_post_age, suggest "month"); whether they want recurring runs (frequency, suggest "once").

3) li_posts_feed — search posts by keywords

Searches LinkedIn for posts matching specific keywords.

URL input: content search URL (/search/results/content).

Ask the user: same as li_posts_monitor — max_post_age and frequency.

4) salesnav_monitor — LinkedIn people search

Searches for people via Sales Navigator, normal LinkedIn search, or event attendees.

URL input: one of:

• Sales Navigator — /sales/search/people
• Normal LinkedIn people search — /search/results/people (no eventAttending)
• Event attendees — contains eventAttending

LinkedIn account selection

When required:

1. mcp__leadspicker__list_linkedin_accounts().
2. Filter:
   • Always exclude is_revoked: True.
   • Sales Nav URLs → keep only sales_nav_session_connected: True.
   • Event URLs → keep only linkedin_session_connected: True.
3. Show a table of valid accounts and ask the user to pick.
4. Use the selected id as linkedin_account_id.
5. If no valid accounts exist, stop and tell the user:
   • Sales Nav: "No LinkedIn accounts with Sales Navigator connected. Please connect one in Leadspicker first."
   • Events: "No connected LinkedIn accounts found. Please connect one in Leadspicker first."

Robot name

Auto-generate the robot's name with a timestamp:

{Type Label} ({DD. M. YYYY HH:MM:SS})

Use mcp__leadspicker__get_current_time(timezone="Europe/Prague") for the timestamp.

Examples: Post Tracker (23. 3. 2026 17:45:28), LinkedIn people (23. 3. 2026 18:10:02).

Workflow

Step 1 — collect input

Resolve target_project_id and the URL(s).

Step 2 — detect robot type

Apply URL detection. Tell the user:

Detected: Post Scrape (li_posts_extract) — scraping likers and commenters from 2 LinkedIn posts.

Step 3 — collect type-specific parameters

• li_posts_extract: none — proceed with defaults.
• li_posts_monitor: ask max_post_age (default "month"), frequency (default "once").
• li_posts_feed: ask max_post_age (default "week"), frequency (default "once").
• salesnav_monitor: if Sales Nav or event URL → run LinkedIn account selection above; ask about frequency if relevant (default "once").

Step 4 — show payload preview

Assemble the parameters and show them to the user for approval:

Robot ready to launch:

Type:           Post Scrape (li_posts_extract)
Name:           Post Tracker (23. 3. 2026 17:45:28)
Target project: CEOs Czechia (#30877)
URLs:           2 LinkedIn posts
Max likers:     250
Max commenters: 250
Frequency:      once

Proceed?

For salesnav_monitor also show: linkedin_account_id (account name) and max_posts (= max people).

Step 5 — launch

mcp__leadspicker__create_robot(
    target_project_id=30877,
    robot_type="li_posts_extract",
    robot_urls="https://www.linkedin.com/feed/update/urn:li:activity:7441864955731103744/",
    frequency="once",
    max_posts=500,
    max_likers=250,
    max_commenters=250,
    should_scrape_authors=False,
    deduplicate_results=True,
    status="ready",
)

For people search with Sales Nav:

mcp__leadspicker__create_robot(
    target_project_id=30877,
    robot_type="salesnav_monitor",
    robot_urls="https://www.linkedin.com/sales/search/people?query=…",
    linkedin_account_id=4773,
    frequency="once",
    max_posts=500,
    max_likers=0,
    max_commenters=0,
    max_post_age="all",
    should_scrape_authors=True,
    deduplicate_results=True,
    status="ready",
)

Step 6 — confirm result

Report:

• Robot ID, name, type, target project.
• If recurring, note the frequency.
• On error, show HTTP status and resolution.

Managing existing robots

Common errors

Safety

• Always show the full payload to the user before launching.
• Never execute without explicit user confirmation.