MCP Server Setup

The Linkt MCP (Model Context Protocol) server enables AI assistants to interact with your Linkt workspace directly. Use natural language to manage ICPs, sheets, tasks, entities, and more.

What is MCP?

MCP (Model Context Protocol) is an open protocol that allows AI assistants to securely connect to external tools and data sources. The Linkt MCP server exposes your workspace through a standardized interface that AI clients can use.

Benefits:

Natural language workflows — Ask Claude to "find all companies in my SaaS ICP" instead of writing API calls
Context-aware assistance — AI understands your workspace structure and can make intelligent suggestions
Automation — Chain operations together conversationally

Configuration

The Linkt MCP server uses HTTP-based transport for reliable, low-latency communication.

Claude Code (CLI)

Add the Linkt MCP server to your project's .mcp.json file (recommended) or global settings:

Project-level (recommended): .mcp.json in your project root

{
  "mcpServers": {
    "linkt": {
      "type": "http",
      "url": "https://api.linkt.ai/mcp",
      "headers": {
        "x-api-key": "${LINKT_API_KEY}"
      }
    }
  }
}

Global: ~/.claude/settings.json

{
  "mcpServers": {
    "linkt": {
      "type": "http",
      "url": "https://api.linkt.ai/mcp",
      "headers": {
        "x-api-key": "${LINKT_API_KEY}"
      }
    }
  }
}

Set your API key as an environment variable:

export LINKT_API_KEY="your_api_key_here"

After saving, restart Claude Code to load the server.

Using direnv for project configuration

For project-based configuration, use direnv with a .envrc file containing dotenv. This automatically loads your .env file when entering the project directory—ideal when using the Linkt Cookbook.

Claude Desktop

Add the configuration to your Claude Desktop settings:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "linkt": {
      "type": "http",
      "url": "https://api.linkt.ai/mcp",
      "headers": {
        "x-api-key": "${LINKT_API_KEY}"
      }
    }
  }
}

Restart Claude Desktop after saving.

API Key Security

Store your API key as an environment variable rather than hardcoding it. For production use, consider using a secrets manager.

Available Tools

The MCP server provides tools for managing all major Linkt resources:

Category	Tools	Description
ICPs	`list_icps`, `get_icp`, `create_icp`, `update_icp`, `delete_icp`	Manage Ideal Customer Profiles
Sheets	`list_sheets`, `get_sheet`, `create_sheet`, `update_sheet`, `delete_sheet`	Manage entity storage containers
Tasks	`list_tasks`, `get_task`, `create_task`, `update_task`, `delete_task`, `execute_task`	Configure and run workflows
Entities	`list_entities`, `get_entity`, `search_entities`, `update_entity`, `delete_entity`	Work with discovered companies and contacts
Runs	`list_runs`, `get_run`, `get_run_queue`, `cancel_run`	Monitor workflow execution
Files	`list_files`, `get_file`, `upload_file`, `delete_file`	Manage uploaded documents

Rate Limiting

MCP requests are subject to the same rate limits as direct API calls:

Limit Type	Value
Requests per minute	60
Concurrent executions	5

If you hit rate limits, Claude will receive an error message and can retry after waiting.

Troubleshooting

Server not connecting

Verify API key — Ensure your LINKT_API_KEY is valid and has the required permissions
Check network — Confirm you can reach api.linkt.ai from your machine
Restart the client — Claude Code and Claude Desktop cache MCP configurations

Tools not appearing

Check configuration syntax — JSON must be valid with no trailing commas
Verify package installation — Run npx @anthropic/linkt-mcp@latest --version to confirm
Check logs — Claude Code displays MCP errors in the terminal output

Permission errors

The MCP server uses your API key's permissions. If a tool fails:

Verify your API key has access to the requested resource
Check that you're targeting resources in your organization
Ensure the resource exists (ICP, sheet, etc.)

Slow responses

Large result sets (many entities, long run histories) may take longer to return:

Use filters to narrow results (e.g., status, entity_type, date ranges)
Request smaller page sizes when listing resources
For entity searches, use specific search terms

Security Considerations

API key scope — The MCP server has full access based on your API key permissions
Data exposure — Claude can see any data returned by the API; avoid sensitive ICPs if needed
Audit logging — All MCP operations are logged with your API key for auditing

Shared Workspaces

If you share your workspace with others, be aware that MCP operations appear under your API key in audit logs, regardless of who initiated the prompt.

Next Steps

Agent Automations — Pre-built Claude Code skills for GTM workflows
Authentication — Managing API keys
Quickstart — Build your first workflow
API Reference — Complete endpoint documentation

On this page