PAL Planner - Task Planning

Break down complex tasks through interactive, sequential planning with revision and branching.

When to Use

• Complex project planning
• System design and architecture
• Migration strategies
• Feature implementation planning
• Breaking down large tasks
• Exploring alternative approaches

Quick Start

# Start planning
result = mcp__pal__planner(
    step="""
    Task: Implement user authentication system

    Scope:
    - Email/password login
    - OAuth (Google, GitHub)
    - JWT tokens with refresh
    - Password reset flow
    """,
    step_number=1,
    total_steps=4,
    next_step_required=True
)

# Continue planning
result = mcp__pal__planner(
    step="Step 2: Database schema and models",
    step_number=2,
    total_steps=4,
    next_step_required=True,
    continuation_id=result["continuation_id"]
)

Required Parameters

| Parameter | Type | Description | |-----------|------|-------------| | step | string | Planning content for this step | | step_number | int | Current step | | total_steps | int | Estimated total steps | | next_step_required | bool | More planning needed? |

Optional Parameters

| Parameter | Type | Description | |-----------|------|-------------| | is_step_revision | bool | Revising a previous step? | | revises_step_number | int | Which step being revised | | is_branch_point | bool | Creating an alternative branch? | | branch_id | string | Name for this branch | | branch_from_step | int | Step this branch starts from | | more_steps_needed | bool | Need more steps than estimated? | | continuation_id | string | Continue session | | model | string | Override model |

Branching for Alternatives

Explore different approaches:

# Create a branch for alternative approach
mcp__pal__planner(
    step="Alternative: Use session-based auth instead of JWT",
    step_number=3,
    total_steps=5,
    next_step_required=True,
    is_branch_point=True,
    branch_id="session-auth",
    branch_from_step=2,
    continuation_id=result["continuation_id"]
)

Revising Steps

Update earlier decisions:

mcp__pal__planner(
    step="Revised: Add rate limiting to auth endpoints",
    step_number=2,
    total_steps=4,
    next_step_required=True,
    is_step_revision=True,
    revises_step_number=2,
    continuation_id=result["continuation_id"]
)

Planning Structure

Step 1: Define Task & Scope

step="""
Task: [What needs to be done]

Scope:
- [Requirement 1]
- [Requirement 2]

Constraints:
- [Timeline, resources, etc.]

Success Criteria:
- [How we know it's done]
"""

Subsequent Steps

step="""
Phase 2: [Phase Name]

Tasks:
1. [Specific task]
2. [Specific task]

Dependencies:
- Requires: [Previous phase]
- Blocks: [Next phase]

Risks:
- [Potential issues]
"""

Example: Migration Planning

# Step 1: Define migration
mcp__pal__planner(
    step="""
    Task: Migrate from PostgreSQL to MongoDB

    Scope:
    - User data (500K records)
    - Order history (2M records)
    - Product catalog (50K records)

    Constraints:
    - Maximum 4 hours downtime
    - Must maintain data integrity
    - Rollback capability required
    """,
    step_number=1,
    total_steps=6,
    next_step_required=True
)

# Step 2: Data mapping
mcp__pal__planner(
    step="""
    Phase 2: Schema Mapping

    Transformations:
    - users table → users collection (denormalize addresses)
    - orders table → orders collection (embed line items)
    - products table → products collection (embed categories)

    Challenges:
    - Many-to-many relationships need restructuring
    - Foreign keys → document references
    """,
    step_number=2,
    total_steps=6,
    next_step_required=True,
    continuation_id=result["continuation_id"]
)

Best Practices

1. Start with scope - Clear boundaries prevent scope creep
2. Identify dependencies - Know what blocks what
3. Plan for rollback - Every step should be reversible
4. Use branches - Explore alternatives before committing
5. Revise freely - Plans should evolve with understanding
6. Include risks - Anticipate what could go wrong