Looking for MCP workflow testers: Interlock drift detection + audit evidence

[MaazAhmed47]

I built Interlock, an open-source MCP runtime trust layer for AI agents.

The narrow problem it focuses on: approved MCP tools should not stay trusted forever if their schema, data access, external reach, side effects, auth scope, or behavior changes after approval.

Interlock baselines approved tools, detects meaningful post-approval drift, can hold/quarantine risky changes before execution, and records audit evidence through Security Receipts.

I’m looking for 1-2 real non-production MCP workflows to test this against this week.

Best fit:

• MCP server builders
• agent teams using real tool access
• DevTools / AI infra builders
• security engineers evaluating MCP risk

For anyone open to testing, I’ll help set it up, review the workflow with you, and share a short drift/quarantine/audit report from the test. If it exposes a real blocker, I’ll prioritize fixing it. Optional public credit as an early design partner, only if you want that.

Website: https://getinterlock.dev
GitHub: https://github.com/MaazAhmed47/Interlock
Demo: https://youtu.be/zYDgD8Eo7uc

Would love feedback from mcp-use builders on where this boundary should live: client, framework, gateway, or MCP server signal.