Context Cache Invalidation Detector

Context Cache Invalidation Detector MCP Connector for Claude

A+

Detects context loss and reasoning drift caused by aggressive prompt caching.

3 tools Official Updated Oct 1, 2026 Official Vinkius Partner

This MCP server provides diagnostic tools to identify when prompt caching or session management has caused the loss of critical reasoning, system instructions, or conversational continuity. It uses deterministic SHA-256 checksumming to monitor the 'Four Pillars of Context': System Prompt, CLAUDE_md Content, Conversation Summary, and the Last Reasoning Block. By comparing current checksums against expected states, it can detect if an AI agent's reasoning has been prematurely cleared or if instructions have been truncated. Use check_context_integrity to validate session health and get_continuity_status to quickly assess if a session is Healthy, Degraded, or Broken.

checksumcontext-lossprompt-cachingreasoning-integritysession-management

3 tools expose this connector's capabilities to your AI agent.

compute_pillar_checksums

Generates unique fingerprints for the four pillars of context

get_continuity_status

Assess session health based on continuity score

check_context_integrity

Determines if the current session context is intact

See how to talk to your AI agent using Context Cache Invalidation Detector.

Check if my current session context is still valid.

The session is Healthy with a continuity score of 1.0. All context pillars are intact.

What is the current status of my session health?

The session status is Degraded. The continuity score is 0.75, indicating one pillar has drifted.

Generate checksums for my current context.

{"systemPromptHash": "a1b2c3d4...", "mdContentHash": "e5f6g7h8...", "conversationSummaryHash": "i9j0k1l2...", "lastReasoningHash": "m3n4o5p6..."}

The `context_continuity_score` represents the health of the session by calculating the ratio of matched checksums against the expected Four Pillars of Context.

Related Connectors