Magenta Canon is a verifiable MCP safety gateway. It sits between an AI agent — Claude Code, Cursor, or your own — and the tools it can call, gating each action against operator-delegated capabilities before it executes, recording every allow and block, and producing cryptographic evidence anyone can verify with a standalone verifier.
- Gate-first: authorized actions are allowed and over-authority actions are blocked before they reach the tool
- Recorded: every allow and block becomes a signed, hash-chained receipt on a Merkle transparency log
- Verifiable: a standalone verifier re-derives the cryptography and shares no code with the server
VERIFIABLE MCP SAFETY GATEWAY
Govern an AI agent's actions before they execute — and prove it after.
Put Magenta between an AI agent and the tools it can call. It allows authorized actions, blocks the rest before they reach the tool, records both, and produces evidence anyone can verify.
A single-tenant reference implementation you run yourself — verify, don't trust.
What are you responsible for?
Pick your role — get the questions you're going to ask, answered, with proof.
CEO / Founder
Your agents act on real systems. Can you prove they're under control?
CISO / Security Leader
Default-deny for agent actions — with tamper-evident receipts.
VP Engineering / CTO
Ship agent features without handing agents the keys.
AI Platform Lead
One gate for every team's agents — capability-scoped, evidenced.
Legal / Compliance
A non-repudiable record of what AI agents did — and didn't do.
Technical Evaluator
Clone, run, verify, tamper. Judge it from source.
Procurement / Risk
Judge the claims against the evidence — not the deck.
Investor / Partner
A verifiable-control wedge for the agent era — under-claimed on purpose.
What Magenta Canon Is
Magenta Canon is a verifiable MCP safety gateway for AI-agent tool calls. It evaluates each action against the operator-delegated capability at the gate — allowing authorized calls, blocking over-authority calls before they reach the downstream tool, and capturing tamper-evident evidence for both.
The Proof Loop
The same demo runs every time: an $89 refund is allowed, a $250 refund is blocked at the gate, the downstream tool's own log shows the blocked call never arrived, the evidence verifies, and one flipped byte makes verification fail.
-
Allowed
An $89 refund within the delegated ceiling is forwarded to the tool.
-
Blocked
A $250 refund exceeds the ceiling and is blocked before it reaches the tool.
-
Recorded
Allow and block alike become signed, hash-chained receipts on a Merkle log.
-
Verified
A standalone verifier returns VERIFIED — and fails if the evidence is tampered.
Verify, Don't Trust
The standalone verifier re-derives the cryptography and shares no code with the server. A committed sample evidence bundle ships in the repository, so the allow and block decisions can be checked independently — without trusting, or running, the Magenta server.
Honest Scope
Magenta Canon is a single-tenant reference implementation you run yourself. The hosted evidence surface is ephemeral unless the durable Postgres witness is activated, and the demo's witness key is not a production key. Under-claiming is the brand: roadmap is labeled as roadmap.