Triple-Gate Validation MCP
Patent-pending Triple-Gate runtime validation — USPTO 64/006,491 commercial embodiment.
AND-logic enforcement · cryptographic proof · MCP tool-callable
Sunaiva Gate validates every agent action before it executes. Start free with 35 constitutional rules running locally — upgrade for all 104 rules, semantic analysis, custom rules, and Ed25519-signed audit evidence. Patent-pending Triple-Gate architecture (USPTO 64/006,491), live today.
The Gate is the free commodity on-ramp — 35 rules, no account, no API key. Triple-Gate is the patent-pending engine underneath, surfaced as its own product: conjunctive AND-logic validation over outputs, with persistent memory and pre-deploy audit layered on the same cryptographic spine.
Constitutional hook engine for AI coding agents. Intercepts every agent action and blocks rule violations before they execute — 104 rules, 35 enforced locally with no account, no API key, no telemetry.
35 local rules · MCP-native · fail-safe · no account required
npm install -g @sunaiva/gate
npx @sunaiva/gate --smoke-test
→ Gate loaded · 35 local rules · HEALTHY
Patent-pending Triple-Gate runtime validation — USPTO 64/006,491 commercial embodiment.
AND-logic enforcement · cryptographic proof · MCP tool-callable
Persistent semantic recall across AI agent sessions.
MCP-native · model-agnostic · team graphs · multi-project
14-layer pre-deploy audit with HMAC-signed verdict.
Exit code 0/1/2 CI gates · Article 12 dossier export · audit hash chain
Rules written into a README, a system prompt, or a style guide are advisory — the agent reads them, then does whatever its sampler decides. Sunaiva moves enforcement out of the prompt and into the protocol: every product here intercepts an action or an output and applies deterministic, mathematical, cryptographically-bound checks before anything ships.
The Gate enforces agent actions. The Triple-Gate MCP enforces model outputs with conjunctive AND-logic. Memory and Ship Confidence extend the same spine to recall and release.
Rules evaluate locally in under a millisecond. Same input, same verdict, every time — no model in the hot path for the free tier.
Local rules enforce fail-closed — errors produce a structured audit entry, never a silent pass. Opt-in fail-open available via env var for evaluation environments.
Speaks the Model Context Protocol directly. Tool-callable by the generating model — validation invoked during generation, not bolted on after.
Triple-Gate and Ship Confidence emit signed verdicts: gate signatures, output hash and timestamp bound into a composite proof.
The Constitutional Gate is free for personal and evaluation use under BUSL-1.1. Paid tiers add server-side rules, multi-seat and custom rule authoring — Pro $49/mo, Team $199/mo (5 seats included · need more? contact us).