Corcava logo Le seul outil métier dont vous avez besoin Corcava
Menu

Serveur MCP gestion des tâches : créer, mettre à jour et prioriser les tâches en toute sécurité

A task management MCP server enables AI assistants to create, read, update, and delete tasks through natural language. This guide covers common workflows and safe prompting patterns for write operations.

Common Task Management Workflows

Task management typically follows four phases:

1. Capture

Quickly record tasks as they come up:

  • Create tasks from conversations
  • Extract tasks from meeting notes
  • Capture ideas and to-dos

Example: "Create a task to follow up with Acme Corp next week"

2. Prioritize

Organize and rank tasks:

  • Update task priorities
  • Set due dates
  • Assign to team members

Example: "Mark the login feature task as high priority"

3. Execute

Work on tasks and track progress:

  • Update task status
  • Add progress comments
  • Start time tracking

Example: "Update the dashboard task status to in progress"

4. Review

Reflect and improve:

  • Review completed tasks
  • Identify blockers
  • Clean up stale tasks

Example: "Show me tasks that haven't been updated in 30 days"

Safe Prompting Patterns for Write Actions

When AI assistants can create, update, or delete tasks, safety is critical. Use these patterns:

⚠️ Always Confirm Before Writing

Never let AI make changes without explicit confirmation. Use these patterns:

  • Preview First: "Show me what task would be created, then I'll confirm"
  • Confirmation Token: "Create this task only if I say CONFIRM"
  • Explicit Approval: "Ask me before creating, updating, or deleting anything"

Pattern 1: Read-First, Then Act

Safe Prompt: "Show me tasks due this week, then help me prioritize them"

AI Behavior:

  1. First: Calls list_tasks (read-only)
  2. Analyzes and presents results
  3. Asks for confirmation before any updates
  4. Only updates after explicit approval

Pattern 2: Preview Before Create

Safe Prompt: "I need to follow up with Acme Corp. Show me what task you'd create, then create it if I approve"

AI Behavior:

  1. Proposes task details (title, due date, project)
  2. Shows preview: "I'll create: [task details]"
  3. Waits for confirmation
  4. Calls create_task only after approval

Pattern 3: Diff Preview for Updates

Safe Prompt: "Update the login task status to done, but show me what will change first"

AI Behavior:

  1. First: Calls get_task to read current state
  2. Shows diff: "Current: [status], New: [status]"
  3. Asks for confirmation
  4. Calls update_task only after approval

Pattern 4: Confirmation Token

Safe Prompt: "Create a task to review the API design. Only create it if I type CONFIRM"

AI Behavior:

  1. Shows what will be created
  2. Asks: "Type CONFIRM to create this task"
  3. Waits for exact token match
  4. Creates task only if token matches

Task Management Tools

Corcava provides comprehensive task management tools:

list_tasks

Purpose: Search and filter tasks

Safety: Read-only, safe to use

Use Cases: Finding tasks, weekly planning, status checks

Full reference →

get_task

Purpose: Get detailed information about a single task

Safety: Read-only, safe to use

Use Cases: Reviewing task details, preparing updates

Full reference →

create_task

Purpose: Create new tasks

Safety: ⚠️ Write operation - requires confirmation

Use Cases: Capturing new work, creating follow-ups

Full reference →

update_task

Purpose: Update task fields (status, priority, due date, etc.)

Safety: ⚠️ Write operation - requires confirmation

Use Cases: Changing status, updating priorities, assigning owners

Full reference →

delete_task

Purpose: Delete tasks (irreversible)

Safety: ⚠️⚠️ Destructive operation - requires explicit confirmation

Use Cases: Cleaning up duplicates, removing obsolete tasks

Full reference →

Example Workflows

Workflow: Capture from Conversation

You: "I need to review the API design doc and get feedback from the team"

AI (Safe Pattern):

  1. Proposes: "I'll create a task: 'Review API design doc and get team feedback'"
  2. Shows: "Project: [suggested], Due: [suggested]"
  3. Asks: "Should I create this task? (Type YES to confirm)"
  4. Creates only after confirmation

Workflow: Prioritize Tasks

You: "Show me high-priority tasks and help me reorder them"

AI (Safe Pattern):

  1. Calls list_tasks filtered by priority
  2. Presents list with current priorities
  3. Suggests new order
  4. Asks: "Should I update priorities? (Type CONFIRM)"
  5. Updates only after confirmation

Workflow: Update Status

You: "Mark the login feature task as done"

AI (Safe Pattern):

  1. Finds task using list_tasks or get_task
  2. Shows: "Current status: In Progress → New status: Done"
  3. Asks: "Update status? (Type YES)"
  4. Updates only after confirmation

Safety Troubleshooting

If you're concerned about accidental writes, check these resources:

Best Practices

Task Management Best Practices

  • ✅ Always preview before creating or updating
  • ✅ Use confirmation tokens for write operations
  • ✅ Read first, then act (get_task before update_task)
  • ✅ Be specific with task titles and descriptions
  • ✅ Use filters when listing tasks (avoid large lists)
  • ✅ Review before deleting (always confirm deletions)
  • ✅ Add comments to explain changes
  • ✅ Use projects and boards for organization

Manage Tasks with AI Assistants

Create, update, and prioritize tasks safely through natural language

Aucune carte bancaire requise