io.github.linkedmash/linkedmash MCP Server

This MCP server enables a workflow centered on LinkedIn saved posts: it imports them into a searchable library, keeps the library synced, and exposes access for external tools. It supports operations such as searching, labeling, and exporting, with integrations mentioned for Notion, Google Sheets, Airtable, and Miro.

🛠️ Key Features

Search, label, and export LinkedIn saved posts
Auto-syncs the saved-post library to Notion, Google Sheets, Airtable, and Miro
Provides a REST API and an MCP server for shared library access

🚀 Use Cases

Let AI agents such as Claude, ChatGPT, Cursor, and Codex search and manage the same saved-post library
Draft, schedule, and publish from imported LinkedIn saved posts

⚡ Developer Benefits

Exposes a “developer surface” via the hosted MCP server, @linkedmash/mcp npm package, and REST API
Enables AI-agent skills to operate on a shared, synced library

⚠️ Limitations

The service source code is not published in this repository

LinkedMash

LinkedMash imports your LinkedIn saved posts into a searchable library and auto-syncs them to Notion, Google Sheets, Airtable, and Miro, with a REST API and an MCP server so AI agents such as Claude, ChatGPT, Cursor, and Codex can search and manage the same library.

This repository is the public home of the LinkedMash developer surface: the hosted MCP server, the @linkedmash/mcp npm package, the REST API, and the agent skill. The product itself lives at linkedmash.com. The service source code is not published in this repository.

Connect an AI agent over MCP

LinkedMash runs a hosted MCP server, so there is nothing to install. Add it to your MCP client config:

json

{
  "mcpServers": {
    "linkedmash": {
      "url": "https://mcp.linkedmash.com/api/mcp?token=<your-token>"
    }
  }
}

Or send the token as a header instead of a URL parameter:

text

https://mcp.linkedmash.com/api/mcp
Authorization: Bearer <your-token>

Prefer the header form where your client supports it; the URL form exists for clients that cannot set headers, and query strings can end up in request logs.

Create a token at linkedmash.com/integrations/api (login required). Note: the server accepts the initial MCP handshake before checking your token, so a wrong token shows up as a failure on the first tool call, not at connect time.

Per-client setup for Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, Codex, OpenClaw, and ChatGPT: Connect Your Agent.

Local MCP server (npm)

Prefer a local stdio server? The same tools ship as @linkedmash/mcp:

bash

LINKEDMASH_TOKEN=<your-token> npx -y @linkedmash/mcp

REST API

The REST API lives at https://api.linkedmash.com/v1 with Bearer authentication (saved posts are called bookmarks in the API):

bash

curl -H "Authorization: Bearer <your-token>" "https://api.linkedmash.com/v1/bookmarks"

Start here:

Agent-readable Markdown versions exist for most doc pages: append .md to an api-docs URL, or start from llms.txt.

Agent skill

Teach a coding agent to drive LinkedMash end to end:

bash

npx skills add linkedmash/agent-skills

Skill repo: linkedmash/agent-skills · Docs: Skills

What LinkedMash does

Imports the posts you save on LinkedIn, text and media included, via a Chrome extension
Full-text and AI search, labels, and smart folders across your whole library
Auto-sync to Notion, Google Sheets, Airtable, and Miro
Download your saved posts as CSV, JSON, or PDF
An AI assistant that answers questions across your saves
A Composer to draft, schedule, auto-publish, and analyze LinkedIn posts

Support

Questions or problems with the MCP server or API? Open an issue or reach out via linkedmash.com.

is_unread_only	boolean	optional	Filter to ONLY unread bookmarks. OMIT unless the user explicitly asks for unread posts. Default false: read AND unread bookmarks are both included.
hide_archived	any	optional	Hide archived bookmarks. Default true (archived excluded). Set false ONLY when the user explicitly asks to include archived posts — do not fill this in otherwise.
media_type	string	optional	Optional media-attachment filter. Accepted values: 'none', 'image', 'video', 'pdf'. When the user did NOT explicitly name a kind of media, pass 'none' (or omit the field) — 'none' means no media filter. Only pass image/video/pdf when the user's own words asked for that kind.
author	string	optional	Filter by author name or ID. Use this parameter when the user wants to find posts by a specific author (e.g., 'Show posts by John Doe', 'Find posts from Suresh sambantham'). Supports fuzzy matching - you can provide the full name, partial name, or exact author ID (e.g., 'urn:li:member:123'). The system will automatically resolve author names to IDs.
tag	string	optional	Filter by tag
posted_from	string	optional	Filter by post posting date range (ISO date string)
posted_to	string	optional	Filter by post posting date range (ISO date string)
bookmarked_from	string	optional	Filter by bookmark date range (ISO date string)
bookmarked_to	string	optional	Filter by bookmark date range (ISO date string)
comment	string	optional	Filter by comment: 'only' returns just saved comments, 'exclude' returns only top-level posts. If the user didn't ask about saved comments, prefer 'none' (or omit) — 'none' applies no comment filter. Paid plans.
limit	number	optional	Number of bookmarks to retrieve (1-100)
cursor	string	optional	Pagination cursor for next page
sort_by	string	optional	Sort order for results

is_unread_only	boolean	optional	Filter to ONLY unread bookmarks. OMIT unless the user explicitly asks for unread posts. Default false: read AND unread bookmarks are both included.
hide_archived	any	optional	Hide archived bookmarks. Default true (archived excluded). Set false ONLY when the user explicitly asks to include archived posts — do not fill this in otherwise.
media_type	string	optional	Optional media-attachment filter. Accepted values: 'none', 'image', 'video', 'pdf'. When the user did NOT explicitly name a kind of media, pass 'none' (or omit the field) — 'none' means no media filter. Only pass image/video/pdf when the user's own words asked for that kind.
author	string	optional	Filter by author name or ID. Use this parameter when the user wants to find posts by a specific author (e.g., 'Show posts by John Doe', 'Find posts from Suresh sambantham'). Supports fuzzy matching - you can provide the full name, partial name, or exact author ID (e.g., 'urn:li:member:123'). The system will automatically resolve author names to IDs.
tag	string	optional	Filter by tag
posted_from	string	optional	Filter by post posting date range (ISO date string)
posted_to	string	optional	Filter by post posting date range (ISO date string)
bookmarked_from	string	optional	Filter by bookmark date range (ISO date string)
bookmarked_to	string	optional	Filter by bookmark date range (ISO date string)
comment	string	optional	Filter by comment: 'only' returns just saved comments, 'exclude' returns only top-level posts. If the user didn't ask about saved comments, prefer 'none' (or omit) — 'none' applies no comment filter. Paid plans.
q	string	optional	Keyword or text search query for finding bookmarks by exact words or phrases in post content. Do NOT use this for author names - use the 'author' parameter instead.
vector_search_term	string	optional	Semantic search query for retrieving bookmarks based on meaning and context, not just keywords. Use for natural language questions or topic-based searches.
limit	number	optional	Number of bookmarks to retrieve (1-100)
cursor	string	optional	Pagination cursor for next page
sort_by	string	optional	Sort order for results

is_unread_only	boolean	optional	Filter to ONLY unread bookmarks. OMIT unless the user explicitly asks for unread posts. Default false: read AND unread bookmarks are both included.
hide_archived	any	optional	Hide archived bookmarks. Default true (archived excluded). Set false ONLY when the user explicitly asks to include archived posts — do not fill this in otherwise.
media_type	string	optional	Optional media-attachment filter. Accepted values: 'none', 'image', 'video', 'pdf'. When the user did NOT explicitly name a kind of media, pass 'none' (or omit the field) — 'none' means no media filter. Only pass image/video/pdf when the user's own words asked for that kind.
author	string	optional	Filter by author name or ID. Use this parameter when the user wants to find posts by a specific author (e.g., 'Show posts by John Doe', 'Find posts from Suresh sambantham'). Supports fuzzy matching - you can provide the full name, partial name, or exact author ID (e.g., 'urn:li:member:123'). The system will automatically resolve author names to IDs.
tag	string	optional	Filter by tag
posted_from	string	optional	Filter by post posting date range (ISO date string)
posted_to	string	optional	Filter by post posting date range (ISO date string)
bookmarked_from	string	optional	Filter by bookmark date range (ISO date string)
bookmarked_to	string	optional	Filter by bookmark date range (ISO date string)
comment	string	optional	Filter by comment: 'only' returns just saved comments, 'exclude' returns only top-level posts. If the user didn't ask about saved comments, prefer 'none' (or omit) — 'none' applies no comment filter. Paid plans.

post_ids	array	required	Specific post IDs to label — a small, explicitly chosen set. For 'all posts matching a filter' use tag_bookmarks_by_filter instead (paginated ids cover only one page).
label_id	string	optional	ID of the existing label to add (use either label_id or label_name)
label_name	string	optional	Name of the label to add (will create if doesn't exist, use either label_id or label_name)

post_ids	array	required	Array of post IDs to remove labels from
label_id	string	optional	ID of the label to remove (use either label_id or label_name)
label_name	string	optional	Name of the label to remove (use either label_id or label_name)

content	string	required	The body text of the LinkedIn post/draft
title	string	optional	Optional internal title for the draft
format	string	optional	Post format. OMIT (or pass 'none') unless the user asked for a specific format — default is text.
post_type	string	optional	Attachment type — set to image/video/pdf ONLY when actually attaching media (it drives how the post publishes to LinkedIn). No media attached -> prefer 'none' or omit.
media_url	string	optional	Public URL of an image/video/PDF to attach (host one with upload_media). Set post_type to match.
media_urls	array	optional	Optional array of media URLs to attach
media_mime_type	string	optional	MIME type of the attached media, e.g. image/png, video/mp4, application/pdf

id	string	required	The id of the content post/draft to update
content	string	optional	Updated body text (required for non-trash updates)
title	string	optional	Updated internal title
format	string	optional	Updated post format. OMIT (or pass 'none') unless the user asked to change it.
media_urls	array	optional	Updated media URLs
media_mime_type	string	optional	Updated media MIME type

url	string	optional	Public http(s) URL of the media to host (recommended — it then stays available until a scheduled publish)
file_base64	string	optional	Base64-encoded file bytes (alternative to url; best for smaller files)
file_name	string	optional	File name, e.g. 'chart.png' (required when using file_base64)
mime_type	string	optional	MIME type, e.g. image/png, video/mp4, application/pdf

id	string	required	The id of the draft to schedule or publish
publish_at	string	required	When to publish: 'now', 'next-slot' (next free posting slot), or an ISO-8601 timestamp
timezone	string	optional	IANA timezone, e.g. 'America/New_York'. Required unless publish_at is 'now'.
post_type	string	optional	Post type (default: text)
media_url	string	optional	Optional media URL to attach
media_mime_type	string	optional	Optional media MIME type

count	number	optional	How many free slots to return
timezone	string	optional	IANA timezone for the slots

date_range	string	optional	Time window, e.g. '7d', '30d', '90d' (default 30d)
limit	number	optional	Max number of recently-published posts to return

kind	string	required	What kind of durable memory this is. 'preference' = how the user likes things done; 'style' = the user's writing voice (tone/structure/audience/formats/signature phrases); 'fact' = a stable fact about the user (role, company, niche, timezone); 'instruction' = a standing instruction ('always...', 'never...').
key	string	required	Short stable slug identifying this memory within its kind (e.g. 'tone', 'audience', 'timezone', 'response_length'). Re-using the same key overwrites the previous value.
value	string	required	The memory content — a concise phrase (keep it under ~160 characters).
mode	string	optional	How to write when this (kind, key) already exists: 'overwrite' (default) replaces the value; 'append' adds it on a new line. OMIT (or pass 'none') for the default — only use 'append' to grow an accumulating note instead of replacing it.

key	string	required	The slug (key) of the memory to delete. Removes every memory with that key for the user.
kind	string	optional	Optional: restrict deletion to this kind only. Prefer 'none' (or omit) unless the user named a kind — 'none' removes every kind sharing the key.

slug	string	required	The smart folder's slug (from list_smart_folders).
bucket	string	optional	Tab/stage name within the folder (a triage bucket); omit for all posts in the folder.
limit	number	optional	Number of posts to retrieve (1-100).
cursor	number	optional	Numeric offset for pagination.

name	string	required	Folder name.
description	string	optional	Optional description of the folder.
base_query	string	optional	The folder's filter query.
triage_buckets	array	optional	Tabs/stages inside the folder.

io.github.linkedmash/linkedmash MCP Server

🛠️ Key Features

🚀 Use Cases

⚡ Developer Benefits

⚠️ Limitations

Topics

search_docs

list_bookmarks

search_bookmarks

get_bookmark_count

list_labels

add_labels

remove_labels

tag_bookmarks_by_filter

remove_labels_by_filter

archive_bookmarks_by_filter

delete_bookmarks_by_filter

setup_vector_store

get_bookmark

list_drafts

create_draft

get_draft

update_draft

delete_draft

upload_media

schedule_post

get_queue

get_posting_slots

cancel_scheduled_post

get_posting_schedule

set_posting_schedule

get_analytics

get_post_analytics

get_growth_digest

remember

recall

forget

list_smart_folders

get_smart_folder

get_smart_folder_posts

create_smart_folder

update_smart_folder

delete_smart_folder

trigger_pdf_export

trigger_csv_export

trigger_json_export

LinkedMash

Connect an AI agent over MCP

Local MCP server (npm)

REST API

Agent skill

What LinkedMash does

Links

Support

Install

Configuration

Related servers