MCP server

Linnia runs a remote Model Context Protocol server. Add it to your AI assistant once and the assistant can work on the LinkedIn accounts you allow, while Linnia keeps every action paced, budgeted and, in approve mode, waiting for a person.

Server URL	https://mcp.linnia.ai/mcp
Transport	Streamable HTTP
Sign-in	OAuth 2.1 with dynamic client registration and PKCE. You sign in with your Linnia dashboard email and password; no API key needed.
Scopes	read, write:messages, write:network, write:content, write:notes

Connect your assistant

Claudeclaude.ai and the desktop app: custom connector.

ChatGPTDeveloper mode connector.

Claude CodeOne command in your terminal.

CursorOne-click install or mcp.json.

VS CodeCopilot agent mode.

Codexconfig.toml and codex mcp login.

Other clientsAny client with OAuth 2.1 and DCR.

How a session works

1. Sign in once. When your assistant connects, a Linnia page opens. Sign in with your dashboard email and password (and your two-factor code if you use one). If your workspace has more than one LinkedIn account, tick the accounts this assistant may use.
2. Ask in plain words. The assistant picks the tools. Tools take an optional account; with one allowed account it is used automatically.
3. Reads come back right away. Searches, profiles and the inbox return within the same answer.
4. Writes return a job. Invitations, messages, comments and posts are queued and sent one at a time inside working hours. The assistant calls get_job to follow up, and never needs to resend.
5. People approve. In approve mode, writes wait in Approvals in the dashboard. The assistant can list what is waiting with pending_approvals, but it can’t approve anything.

How long the assistant waits

get_job holds the request open for as long as the client allows: about 170 seconds on Claude.ai, about 590 seconds on Claude Code, Cursor, Codex and VS Code, and about 45 seconds on ChatGPT. If the job is still waiting (for working hours, budget or approval), it returns a snapshot and the assistant can ask again later.

What a connection can reach

• One connection covers the accounts you ticked at sign-in (or every account, if the workspace has only one). Tools refuse other accounts.
• Access tokens last 8 hours and refresh automatically for up to 30 days of inactivity.
• A connection stops working when the person who signed in is disabled, the workspace is suspended, or, for connections made with an API key, the key is revoked. See Troubleshooting.

Tools, prompts and resourcesEverything the assistant can call, with what each one does.