Kondo's LinkedIn MCP server for Claude Code, Codex & other AI agents
Kondo's LinkedIn MCP server lets Claude Code, Codex, ChatGPT and other AI agents read your LinkedIn DMs, draft replies and sort chats into folders.
What is Kondo's LinkedIn MCP server?
It's a hosted MCP (Model Context Protocol) server that gives AI agents access to the LinkedIn inbox you manage in Kondo. Any agent that supports remote MCP servers over Streamable HTTP can use it. It’s designed to work seamlessly with popular AI assistants like Claude, ChatGPT, Cursor, and more.
What can you do with Kondo's LinkedIn MCP server?
- Chat search — Find chats by name or message text, and read whole conversations
- Reply drafting — Draft replies in your chats for you to review and send from Kondo
- Inbox organization — Label chats into folders, archive what's done, and set reminders to follow up
- Connection lookup — Search your connections by name, company, title or school, and keep notes on them
- Invite handling — See pending connection invites, and accept or ignore them
- Snippets — Use and create saved reply templates with smart placeholders
The server reaches your inbox through Kondo running in your browser, so your messages stay in your browser rather than on Kondo's servers. Your agent reads the messages Kondo has already stored first, and only goes to LinkedIn itself when it needs more, one request at a time.
This page is for LinkedIn, which Kondo runs in your browser. For WhatsApp, which runs in the Kondo desktop app, see MCP setup for the desktop app.
Set up Kondo's LinkedIn MCP server
Before you connect
Kondo's LinkedIn MCP server is part of the Business plan. Check your plan in Kondo's settings, and upgrade there or on the pricing page if you need to.
- Sign in to Kondo at app.trykondo.com in Chrome, Edge, Brave or another Chromium browser.
- Add the Kondo browser extension when Kondo asks for it, and wait until your LinkedIn inbox shows in Kondo.
- Keep Kondo open in one browser tab while your agent works. The server only answers while that tab is open.
Connect Kondo to Claude Code
- In your terminal, add Kondo to Claude Code:
- Start Claude Code. If it's already open, quit and start it again so it picks up Kondo.
- Type
/mcp, select kondo, and choose Authenticate. - Your browser opens. Sign in to your Kondo account, then go back to Claude Code:
/mcpnow shows kondo as connected.
--scope user makes Kondo available wherever you open Claude Code, not just in the folder you ran the command in.
Connect Kondo to Claude on the web or the desktop app
- On the web, open claude.ai/settings/integrations and click Add more integrations, then Add custom MCP server. In the macOS or Windows app, open Settings > Connectors (or Customize > Connectors) and click Add Connector.
- Name it Kondo and paste the server URL:
- Click Add and sign in to your Kondo account.
Start a new chat. Kondo's tools appear in the tools menu.
Connect Kondo to Codex
Run these two commands in your terminal:
The second command opens your browser to sign in to Kondo. Or add the server to ~/.codex/config.toml yourself:
Then run codex mcp login kondo, and start a new Codex session.
Connect Kondo to ChatGPT
Requires a ChatGPT Business, Enterprise, or Edu plan. Plus/Pro users can connect via Developer Mode but cannot publish apps to a workspace.
A workspace Admin or Owner first turns on Developer Mode under Settings > Apps > Advanced settings > Developer mode.
- In ChatGPT, go to Settings > Apps > Create.
- Name it Kondo and paste the server URL:
- Click Create and sign in to your Kondo account.
ChatGPT asks you to confirm every action that changes your inbox, such as saving a draft or adding a label.
Publishing to your workspace
Only Admins and Owners can publish apps for the whole workspace.
- Go to Workspace Settings > Apps
- Click Drafts, find Kondo, then click Publish
- Review any safety warnings (Kondo supports write actions like archiving and labeling)
- Once published, all workspace members will see Kondo in their Apps settings
On Business plans, apps cannot be updated after publishing — you must recreate and republish. Enterprise/Edu plans support updates and RBAC controls.
Connect Kondo to Cursor
- Open Cursor Settings > MCP > Add new global MCP server.
- Paste this configuration:
- Save and restart Cursor, then sign in to Kondo when Cursor asks.
Connect Kondo to VS Code (GitHub Copilot)
Create .vscode/mcp.json in your workspace:
Open the Command Palette (Cmd+Shift+P or Ctrl+Shift+P), run MCP: List Servers to start the server, and sign in to Kondo when asked.
Connect Kondo to other AI agents
Kondo's server uses the Streamable HTTP transport at https://relay.trykondo.com/mcp, so any agent that supports remote MCP servers can connect. Most take a JSON configuration:
Some, such as Gemini CLI, use httpUrl in place of url:
If your agent only runs local stdio servers, use mcp-remote as a bridge:
Check that your agent is connected
Ask your agent: "List my Kondo inboxes." It should reply with your LinkedIn inbox and the folders you've made with labels. This only reads, so it changes nothing. If the agent has no Kondo tools or the request times out, see Troubleshooting.
Give your agent the Kondo skill
The Kondo skill teaches your agent to use the tools well: when to read stored messages, when to fetch more from LinkedIn, and how to stay within LinkedIn's limits. Save it as SKILL.md in your agent's skills folder, or paste it into the agent's instructions.
How to use Kondo's LinkedIn MCP server
Automatically draft LinkedIn DMs
Organize LinkedIn DMs into folders
In Kondo, labels are your folders: each label gets its own inbox in the left sidebar, and a chat can sit in several. Your agent can create labels, sort chats into them, archive what's done and set reminders to bring a chat back.
Triage unread LinkedIn messages
Follow up on LinkedIn conversations
Find people in your LinkedIn network
Manage LinkedIn connection invites
Keep CRM notes on LinkedIn contacts
What Kondo's LinkedIn MCP server can and can't do
- It never sends. Your agent saves drafts; only you send messages, from Kondo.
- Some actions reach LinkedIn. Archiving, reminders and invite responses change your LinkedIn inbox. Labels, notes, drafts and snippets stay in Kondo.
- It stays within LinkedIn's limits. Most reads use the messages Kondo has stored. Requests that go to LinkedIn run one at a time, a few seconds apart, so your account doesn't look automated.
- Your messages stay in your browser. Commands and responses pass through Kondo's relay in memory and aren't stored. See the Privacy Policy.
- It needs Kondo open. With no Kondo tab open, or more than one, your agent's requests fail.
Kondo LinkedIn MCP tools
Inbox and chats
Actions
Connections
Invites and snippets
Troubleshooting problems with Kondo's LinkedIn MCP server
Disconnect an AI agent from Kondo
Go to Settings > MCP in Kondo and click Revoke next to the agent. Its access ends immediately.