Quibbly LinkedIn Inbox MCP

npx add-mcp 'https://server.quibbly.co/mcp'

Quibbly MCP

Give Claude, ChatGPT, Cursor or any MCP client your LinkedIn inbox as Quibbly sees it: every synced conversation, who watched your videos and for how long, your notes, tags and follow-ups. Reads come from Quibbly's own copy and never touch LinkedIn. An assistant can draft replies; they land in your inbox queue and only you send them.

Server URL

https://server.quibbly.co/mcp

Included with Solo Pro and every team plan.

Streamable HTTP. OAuth 2.1 with PKCE and dynamic client registration, or a personal token from Settings → Assistants as a Bearer header.

Connect

1. 1 Claude (web or desktop): Settings → Connectors → Add custom connector, paste the URL. Claude opens Quibbly, you approve, done.
2. 2 ChatGPT: Settings → Connectors → Create, paste the URL, approve on the Quibbly page.
3. 3 Cursor: Settings → MCP → Add new server, transport HTTP, the URL. Use OAuth or a personal token.

Claude Code: claude mcp add --transport http quibbly https://server.quibbly.co/mcp

Tools

list_conversations Threads by status: waiting on you, waiting on them, follow-up due, hot, gone quiet.

search_messages Full-text search across 90 days of messages and contacts.

read_conversation A thread with voice-note transcripts, the Quibblys sent in it, notes and follow-ups.

get_contact A person: profile with LinkedIn public identifier, emails and locations seen on their video views, CRM links, tags, notes, follow-ups, every Quibbly sent and its outcome. Look up by name, profile URL or crm_id.

recent_messages A flat feed of new messages across the inbox with message ids, direction and the person's CRM links. Built for syncing: ask for what is unsynced for a system.

link_contact / mark_synced Remember the CRM record a person maps to, and how far each conversation has been synced, so a connected CRM stays in step run after run.

who_watched Viewers of a video or the whole library with watch percent, time and location.

list_follow_ups / set_follow_up What is due, and reminders an assistant sets for you.

add_note / set_tags / list_tags Notes and tags on people and conversations.

list_quick_replies / list_library / get_library_item Your snippets and videos with send, seen and reply counts. Library search covers what you say in each video, and get_library_item returns the full transcript.

find_attachments / upload_attachment Everything that can ride on a draft: videos and voice notes (searchable by transcript), files kept with snippets, files and images from LinkedIn conversations, or a file the assistant uploads.

draft_reply / withdraw_draft / list_drafts Proposed replies: text, an attachment, or both as one message with the text as its caption. They land in your inbox queue; you send, one click per draft.

sync_status / request_sync How fresh the data is, and a nudge for the extension to sync now.

What an assistant cannot do

• Send a LinkedIn message. Drafts wait in your inbox for you.
• Call LinkedIn. Every read is served from Quibbly's synced copy.
• Exceed its limits: reads, writes and drafts are rate limited per token, and every call is listed in your activity log.

Freshness is always visible

Every response includes how old the data is and whether your extension is online, so an assistant can tell you when it is working from a stale copy.