FAQ MCP pour la gestion de projet : questions courantes sur le protocole MCP
Answers to frequently asked questions about using Model Context Protocol (MCP) for project management. Short answers with links to detailed guides for deeper information.
Setup and Configuration
What is Model Context Protocol (MCP)?
MCP is a protocol that connects AI assistants to external tools and data sources. For project management, it lets your AI assistant interact with task management systems like Corcava. En savoir plus sur MCP →
Which AI assistants support MCP?
Claude Desktop, Cursor, Windsurf, and Continue all support MCP. Each has its own setup process. See client setup guides →
How do I set up Corcava MCP?
Get your API key from Corcava, add the MCP server to your client config, and restart. 5-minute quickstart →
Where do I put the MCP config file?
Location varies by client: Claude Desktop uses ~/Library/Application Support/Claude/claude_desktop_config.json, Cursor uses ~/.cursor/mcp.json. See config generator →
Usage and Workflows
What can I do with MCP for project management?
List tasks, create tasks, update status, add comments, track time, generate reports, and more—all through natural language prompts. See use cases →
How do I create a task using MCP?
Ask your AI assistant: "Create a task in Corcava with title '[Title]' and project '[Project ID]'. Show me a preview before creating." See prompt pack →
Can MCP update tasks automatically?
MCP can update tasks, but you should always require confirmation. Use prompts that show previews and require approval tokens. Approval patterns →
How do I generate weekly status reports?
Ask: "Generate a weekly status report from Corcava tasks: completed work, in progress, blocked items, next week priorities." Status report guide →
Security and Permissions
Is MCP secure for project management?
Yes, with proper setup: use API keys with least privilege, require approvals for writes, and enable audit logging. Security guide →
How do I manage API keys for my team?
Create separate keys per user, name them descriptively, rotate every 90 days, and revoke compromised keys immediately. API key management →
Can I restrict what MCP can do?
Yes, use role-based access: give read-only keys to viewers, contributor keys for day-to-day work, admin keys only for managers. RBAC guide →
Troubleshooting
MCP server not showing up in my client
Check config file location, validate JSON syntax, ensure client was fully restarted. Debugging checklist →
Getting 401 Unauthorized errors
Verify API key is active, check Authorization header format (must be "Bearer YOUR_KEY" with space), ensure key hasn't been revoked. 401 troubleshooting →
Tool calls are slow or timing out
Use pagination, add filters to reduce data, batch operations, and avoid requesting all tasks at once. Pagination guide →
Best Practices
What are the best practices for MCP prompts?
Read before writing, show previews, require confirmation tokens, batch operations, and use filters. Prompting patterns →
How do I roll out MCP to my team?
Start with a pilot group, provide training, establish safe defaults, manage keys properly, and track success metrics. Team rollout guide →
Related Resources
What is MCP?
Introduction to MCP
Quickstart
5-minute setup
Debugging Checklist
Fix common issues
Troubleshooting Index
Complete troubleshooting
Get Started with MCP
Find answers to common questions and get help with MCP for project management