HERMX ProofGate
Enables approval-gated actions for Alexa+ by exposing tools to plan, approve, execute, and verify bounded changes to sandbox settings, including exact approval phrases, precondition checks, drift protection, and SHA-256 verification evidence.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@HERMX ProofGatePlan a change to set the sandbox mode to maintenance"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
HERMX ProofGate
Approval-gated actions for Alexa+ over Streamable HTTP MCP.
ProofGate is a clean-room hackathon demo built for the Build, Ship, Shape: Amazon Developer Hackathon. It demonstrates a safety-first agentic workflow without publishing HERMX commercial internals.
What it demonstrates
An assistant cannot jump directly from intent to effect. Every change follows:
plan → explicit approval → drift-safe execution → cryptographic verification
The public demo exposes five MCP tools:
proofgate_statusplan_changeapprove_changeexecute_changeverify_change
The effect surface is intentionally bounded to two sandbox settings. No shell, customer data, credential path, production API, provider registry, routing logic, tunnel configuration, or proprietary adapter is included.
Related MCP server: agent-mcp-workflow-platform
Why this matters for Alexa+
Voice and conversational agents become more useful when they can take actions, but higher agency also increases the cost of an incorrect action. ProofGate turns approval and verification into protocol-visible steps rather than hidden application behavior.
A typical interaction is:
Alexa+ asks ProofGate to plan a bounded change.
ProofGate returns the before/after state and an exact approval phrase.
The user explicitly approves that exact plan.
ProofGate re-checks preconditions before execution.
ProofGate verifies the result and returns a SHA-256 evidence digest.
Run locally
Requirements: Python 3.11+.
python -m venv .venv
# Windows: .venv\Scripts\activate
# macOS/Linux: source .venv/bin/activate
pip install -e .
python server.pyThe Streamable HTTP endpoint is:
http://127.0.0.1:8767/mcp
The project pins mcp==1.27.2 and uses Streamable HTTP with JSON responses and stateless HTTP sessions.
Test
pip install -e ".[dev]"
pytest -qThe tests prove that execution is blocked without approval, approval phrases must match exactly, state drift blocks stale plans, allowlists constrain the effect surface, and successful execution produces verifiable evidence.
Commercial boundary
This repository is intentionally standalone. It reimplements only the public safety pattern needed for the hackathon demo. It does not contain or depend on the proprietary HERMX control plane.
See SECURITY.md for the disclosure boundary and DEVPOST_SUBMISSION.md for the submission copy.
This server cannot be deployed
Maintenance
Related MCP Connectors
Preflight, approve, and prove consequential agent actions with signed evidence and x402 tools.
Fail-closed action authorization, MCP risk scanning, x402 checks, and signed receipts.
Runtime permission, approval, and audit layer for AI agent tool execution.
Remote MCP for Copilot CLI switch gate MCP, structured receipts, audit logs, and reviewer-ready evid
Related MCP Servers
- FlicenseNot gradedqualityCmaintenanceAn MCP approval gate that ensures agents execute only human-approved, hash-verified actions, providing fail-closed consent for consequential operations.-
- FlicenseNot gradedqualityCmaintenanceEnables approval-gated incident response workflows that gather evidence through read-only MCP tools, perform idempotent writes, and preserve a durable audit trail.-
- AlicenseNot gradedqualityBmaintenanceEnables transparent MCP proxying with a hash-chained effect ledger, classifying agent actions by reversibility, enforcing approval gates, and dry-run previews of sessions.MIT
- AlicenseNot gradedqualityBmaintenanceEnables smart home automation via MCP with human-in-the-loop approval gates, providing device discovery, state queries, policy checks, and proposal-based execution of actions and scenes that only proceed after explicit user confirmation, with full audit logging.MIT