A2A Matrix (A2AMatrix): Governed enterprise AI agent runtime in Rust with sandboxed Claude Code Foundry, MCP hub, human approvals & cryptogr

Hi everyone! I'm Hari Krishna, creator of A2A Matrix (A2AMatrix) — https://a2amatrix.com

Why I built it Giving an AI agent a loop and tools is easy; running it in regulated enterprises without standing authority is the hard part. Most frameworks leave permissions, human approvals, policy testing, drift assurance, regional terminology, and tamper-evident audit trails for your team to build from scratch.

What it does A2A Matrix separates model capability from authority inside a 17-service Rust runtime:

1. Governance & Policy Replay: Models only propose intents; permissions, policy, and human approval gates decide execution, with historical policy replay and an Ed25519-anchored hash chain.
2. Intelligence, AI Assurance & Backtesting: Replay historical decisions against new models or prompts, enforce prompt-regression thresholds, score runs with review rubrics, and trigger governed self-healing / forward-fix on drift.
3. Region, Terminology & Grounded Math: Per-(industry, region) vocabulary overlays, regulatory packs, and sandboxed Wasmtime Compute → Ground → Verify execution.
4. Foundry, Copilot & Blueprint: Paste your website or docs to generate your company's blueprint, compose in Foundry (sandboxed Claude Code CLI), and debug live in Copilot.

What feedback I'd love Sign up free at https://a2amatrix.com (Google, Microsoft, or email + 1-click free OpenRouter models) and let me know what you think!