Kylas CRM MCP Server is a Python-based MCP server that exposes lead-related operations for Kylas CRM. It provides a set of actions you can invoke from any MCP client (for example Cursor or Claude Desktop) to create and manage leads, search and filter leads, and look up related entities like users, products, and pipelines. The server communicates via the MCP stdio transport, meaning clients connect through standard input and output streams. Key capabilities include creating leads with dynamic fields derived from user context, searching leads with multiple criteria, resolving user names to IDs for owners or created-by fields, and translating product names or pipeline names into their corresponding IDs. It also offers utilities to fetch lead field schemas (including custom fields and picklist IDs) and to identify pipeline stages for various lead statuses (open, won, etc.).

Prerequisites:

• Python 3.10 or newer
• pip (usually comes with Python)
• Access to a Kylas account and a valid API key

Installation steps:

1. Create and activate a Python environment (recommended): python -m venv venv source venv/bin/activate # on Unix/macOS venv\Scripts\activate # on Windows

2. Install the MCP server package in editable mode (from repository or PyPI):

   From repository (local development):

   pip install -e .

   Or from PyPI after publishing:

   pip install kylas-crm-mcp-server

3. Ensure dependencies are installed (dev dependencies optional): pip install -e ".[dev]" # if you plan to run tests locally

4. Set up required environment variables (KYLAS_API_KEY, optional KYLAS_BASE_URL) as described in the Configuration section.

5. Run the server (see Running the server in the README): python -m kylas_crm_mcp

   or during development: python main.py

• The KYLAS_API_KEY must be supplied for the MCP server to authenticate with the Kylas API. Exposing this key in logs or environment files without protection is a security risk.
• KYLAS_BASE_URL is optional and defaults to https://api.kylas.io/v1 if not provided.
• The server uses stdin/stdout for communication (stdio transport). Ensure your MCP client is configured to connect via the MCP protocol accordingly.
• If running in Docker, you can pass the API key as an environment variable (e.g., -e KYLAS_API_KEY=your_key) and map standard IO as required by your container orchestration.
• When developing locally, you can use python main.py to run from the repository root as noted in the Running the server section.
• This MCP server supports actions like get_lead_field_instructions, create_lead, search_leads, lookup_users, lookup_products, lookup_pipelines, get_pipeline_stages, and search_idle_leads; refer to the MCP client documentation for exact request formats.