pipefy

This MCP server exposes a set of Pipefy integration tools as MCP primitives that an AI agent (via MCP) can call from within an IDE or automation environment. It includes pipe and card related operations such as get_pipe, get_start_form_fields, get_cards, find_cards, get_card, create_card, add_card_comment, update_comment, delete_comment, delete_card, move_card_to_phase, update_card_field, and update_card. The server uses MCP's elicitation to prompt for required fields before actions that need them (for example, collecting necessary card details before creating a new card). You can use these tools to explore a Pipefy workspace, fetch pipe structures, list or search cards, and perform CRUD operations on cards and comments, all while benefiting from safety checks and structured responses that help you reason about changes before they’re applied.

Prerequisites:

• Python 3.12+ installed
• Access to a Pipefy Service Account with the necessary permissions
• uv (Astral UV) installed for dependency management

Installation steps:

1. Clone the repository: git clone https://github.com/gbrlcustodio/pipefy-mcp-server.git cd pipefy-mcp-server

2. Install and prepare dependencies with uv: uv sync

3. Install and run the MCP server locally (example using uv as described in the docs): uv run --directory . pipefy-mcp-server

4. Verify the server starts and is reachable at the configured GraphQL/HTTP endpoints. If you plan to run via MCP, configure your MCP client (e.g., Cursor) with the mcpConfig details shown in this README.

Notes:

• Ensure your Pipefy credentials (OAuth client/secret) are kept secure and not committed to source control.
• The server may require you to provide additional Pipefy environment details or URL endpoints depending on your deployment environment.

Tips and considerations:

• The server relies on a Pipefy Service Account; keep the credentials secure and rotate them periodically.
• When using delete_card, remember it’s a destructive operation and may require an explicit confirm flag in actual usage depending on client behavior.
• Use elicitation (prompting) features to collect missing fields before performing destructive or bulk updates.
• For local development, you can inspect servers with the MCP Inspector (e.g., npx @modelcontextprotocol/inspector uv --directory . run pipefy-mcp-server).
• If you modify GraphQL schemas or endpoints, run the provided GraphQL schema update commands to keep local tooling in sync.