reason

Capture step-by-step reasoning. Intelligence layer is session-scoped.

Usage

{
  "tool": "reason",
  "arguments": {
    "goal": "Determine root cause of auth failure",
    "steps": [
      "Token is present in request",
      "Token signature validates",
      "Token is expired",
      "Conclusion: token refresh not triggering"
    ]
  }
}

Parameters

Parameter	Type	Required	Description
goal	string	Yes	What you’re reasoning about
steps	string[]	Yes	Sequential reasoning steps
conclusion	string	No	Final conclusion
confidence	float	No	Confidence in conclusion

Examples

Debugging reasoning:

{
  "tool": "reason",
  "arguments": {
    "goal": "Why is the API returning 500?",
    "steps": [
      "Request reaches load balancer (verified in logs)",
      "Load balancer routes to backend (verified)",
      "Backend receives request (verified)",
      "Database query times out (found in traces)",
      "Timeout causes unhandled exception"
    ],
    "conclusion": "Database connection pool exhausted",
    "confidence": 0.85
  }
}

Important

Intelligence is session-scoped. Reasoning chains do not persist across conversations. If you need the conclusion later, also store it via remember or believe.

When to Use

• Complex problem-solving
• Multi-step analysis
• Decision-making with explicit steps
• Debugging investigations

When NOT to Use

• Simple observations (use remember)
• Information you need to persist (add a remember too)