# Linked API MCP server

> Lets AI assistants control a LinkedIn account to search leads, analyze profiles, read conversations, and send messages.

- Repository: https://github.com/Linked-API/linkedapi-mcp
- Page: https://hostedmcp.org/servers/linked-api/linkedapi-mcp
- Transport: stdio, remote
- Stars: 70
- License: MIT
- Latest version: 2.3.21

## Install

### Claude Code

```sh
claude mcp add --env 'IDENTIFICATION_TOKEN=<YOUR_IDENTIFICATION_TOKEN>' --env 'LINKED_API_TOKEN=<YOUR_LINKED_API_TOKEN>' linkedapi-mcp -- npx -y @linkedapi/mcp
```

### Cursor (`~/.cursor/mcp.json`)

```json
{
  "mcpServers": {
    "linkedapi-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@linkedapi/mcp"
      ],
      "env": {
        "IDENTIFICATION_TOKEN": "<YOUR_IDENTIFICATION_TOKEN>",
        "LINKED_API_TOKEN": "<YOUR_LINKED_API_TOKEN>"
      }
    }
  }
}
```

### VS Code (`.vscode/mcp.json`)

```json
{
  "servers": {
    "linkedapi-mcp": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "@linkedapi/mcp"
      ],
      "env": {
        "IDENTIFICATION_TOKEN": "<YOUR_IDENTIFICATION_TOKEN>",
        "LINKED_API_TOKEN": "<YOUR_LINKED_API_TOKEN>"
      }
    }
  }
}
```

### VS Code CLI

```sh
code --add-mcp '{"name":"linkedapi-mcp","type":"stdio","command":"npx","args":["-y","@linkedapi/mcp"],"env":{"IDENTIFICATION_TOKEN":"<YOUR_IDENTIFICATION_TOKEN>","LINKED_API_TOKEN":"<YOUR_LINKED_API_TOKEN>"}}'
```

### Claude Desktop (`claude_desktop_config.json`)

```json
{
  "mcpServers": {
    "linkedapi-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@linkedapi/mcp"
      ],
      "env": {
        "IDENTIFICATION_TOKEN": "<YOUR_IDENTIFICATION_TOKEN>",
        "LINKED_API_TOKEN": "<YOUR_LINKED_API_TOKEN>"
      }
    }
  }
}
```

### Codex

```sh
codex mcp add linkedapi-mcp --env 'IDENTIFICATION_TOKEN=<YOUR_IDENTIFICATION_TOKEN>' --env 'LINKED_API_TOKEN=<YOUR_LINKED_API_TOKEN>' -- npx -y @linkedapi/mcp
```

### Codex config.toml (`~/.codex/config.toml`)

```toml
[mcp_servers.linkedapi-mcp]
command = "npx"
args = ["-y", "@linkedapi/mcp"]

[mcp_servers.linkedapi-mcp.env]
IDENTIFICATION_TOKEN = "<YOUR_IDENTIFICATION_TOKEN>"
LINKED_API_TOKEN = "<YOUR_LINKED_API_TOKEN>"
```

### Gemini CLI

```sh
gemini mcp add -s user -e 'IDENTIFICATION_TOKEN=<YOUR_IDENTIFICATION_TOKEN>' -e 'LINKED_API_TOKEN=<YOUR_LINKED_API_TOKEN>' linkedapi-mcp npx -- -y @linkedapi/mcp
```

### Gemini settings.json (`~/.gemini/settings.json`)

```json
{
  "mcpServers": {
    "linkedapi-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@linkedapi/mcp"
      ],
      "env": {
        "IDENTIFICATION_TOKEN": "<YOUR_IDENTIFICATION_TOKEN>",
        "LINKED_API_TOKEN": "<YOUR_LINKED_API_TOKEN>"
      }
    }
  }
}
```

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `IDENTIFICATION_TOKEN` | yes | yes | Identification token for the LinkedIn account from https://app.linkedapi.io. |
| `LINKED_API_TOKEN` | yes | yes | Linked API access token from https://app.linkedapi.io. |

## FAQ

### What can an AI assistant do with my LinkedIn account through this MCP server?

It can search for leads, analyze profiles, read existing conversations, and send messages and follow-ups. It can also gather company and employee data for market research.

### Why does a tool return a workflowId instead of a result?

Long actions run through a cloud browser and are queued, so the tool returns a status while the work continues. Call get_workflow_result with that workflowId and operationName instead of retrying.

### How do I get an immediate result instead of waiting?

Pass waitSeconds: 0 to get_workflow_result for a single-shot snapshot. Otherwise it long-polls until the workflow completes or the request budget runs out.

### Which AI clients can use this server?

It works with Claude, Cursor, VS Code, and Windsurf. Setup guides are provided for each.
