Feishu Project MCP Work Item Management Playbook

Problem

After connecting Feishu Project MCP to Claude Code, Codex CLI, Cursor, or VSCode Copilot, the friction is usually not “can the API be called?” but repeatedly confirming the space, work item type, field keys, roles, enum options, and linked items. That makes conversations long, expensive, and prone to writing unconfirmed fields.

How It Works

This playbook turns Feishu Project MCP work-item operations into maintainable local context: verify that feishu-project MCP is authorized and connected, then read the business-context file. For creation, call get_workitem_field_meta first, then merge schema-required fields, workflow-required fields, and the fields to write in the current turn. Dynamic values prefer user-provided values, default suggestions, cached candidates, and live queries only when missing, stale, or explicitly requested. After confirmation, call create_workitem, update_field, update_node, or transition_node, and use apply_patch to update new cached values.

Boundaries

It depends on a Feishu Project MCP Server and does not replace backend authorization. Fields, roles, enums, and linked items remain governed by the space configuration. It fits spaces with stable business fields, roles, and defaults, not one-off exploration of unknown projects. Before node transitions, check get_transition_required; a timeout is not automatically a failure, so verify state first.