delve-mcp (v1.0.0)

Force a problem perception check before solving. Surfaces assumptions, explores alternative interpretations, and commits to a justified framing. USE WHEN: - Starting a new problem or task - The problem statement is ambiguous or multi-faceted - You want to verify you're solving the right problem - Multiple stakeholders might define success differently WORKFLOW: 1. State the problem as you understand it 2. List your assumptions with how each could be disproved 3. Provide at least 2 alternative interpretations 4. Choose one and justify why Returns a frame_id to link subsequent reasoning steps.

Record a premise-tracked reasoning step. Each premise declares its own confidence and source, enabling the server to detect contradictions, surface weak assumptions, and score chain stability. USE WHEN: - Working through a multi-step problem - You need to track what you're assuming vs. what you've verified - Building on previous reasoning steps - You want server-side contradiction detection WORKFLOW: 1. Start with step=1 2. Declare premises with per-premise confidence and source (verified/recalled/assumed/derived) 3. State your thought process and outcome 4. Optionally link to a delve-frame result via frame_id 5. Use revises_step to correct earlier reasoning 6. Set is_final_step=true when done RESPONSE INCLUDES: - warnings: unframed reasoning, weak premises, unverified assumptions - contradictions: premises that conflict with earlier steps - stability: chain confidence score and risk level (stable/fragile/critical)

Quick triage: should you use structured reasoning for this task? Analyzes complexity signals and recommends an approach. USE WHEN: - Before starting any task, to decide if full reasoning is warranted - You're unsure if a problem needs structured decomposition - You want to avoid reasoning overhead on simple tasks RETURNS: - "proceed": task is simple, just do it without structured reasoning - "frame_first": task is ambiguous, use delve-frame to clarify the problem - "use_reasoning": task is complex, use delve-frame then delve-reason for full structured reasoning - suggested_depth: estimated number of reasoning steps needed

DELVE server health and configuration status