DanceNitra/inspeximus

Overview

A self-correcting memory layer for AI agents, implemented as a zero-dependency Python library and MCP server. It allows agents to deterministically supersede, revert, or forget stored values, with verifiable erasure and witness-backed tamper-evident receipts. The design targets EU AI Act readiness.

Best for

Best for
Developers building compliant, auditable memory systems for AI agents

Use cases

• Track and revert state changes made by an AI agent
• Provide tamper-evident audit logs for agent memory operations
• Implement compliant data erasure for AI systems

How to use

Install

pip install inspeximus

Tested with

Claude Code, Cursor, Windsurf, Codex, Cline

Notes

A self-correcting memory layer for AI agents, implemented as a zero-dependency Python library and MCP server. It allows agents to deterministically supersede, revert, or forget stored values, with verifiable erasure and witness-backed tamper-evident receipts. The design targets EU AI Act readiness.

6 stars on GitHub. Last updated 2026-09-13. Licensed MIT.

Use cases

• Track and revert state changes made by an AI agent
• Provide tamper-evident audit logs for agent memory operations
• Implement compliant data erasure for AI systems

Pros

• Zero dependencies, easy to integrate into Python projects
• Deterministic operations for reliable state management
• Built-in audit receipts support compliance requirements

Cons

• Very low adoption with 6 GitHub stars, so maturity and community support are unproven
• Niche scope as a memory layer, requiring MCP or Python integration to be useful
• Limited documentation and usage examples due to low project visibility

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.