MCP-native runtime validation for AI agents

Stop documenting rules your AI agent ignores. Enforce them.

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.

Patent-pending validation architecture — see /patents

Conjunctive AND-logic · every action
Identity Gateverify
AND
Safety Gatepolicy + PII
AND
Compliance Gateregulatory
=
ALLOWEDsig + hash + ts

The suite

One validation architecture. Layered by tier.

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.

Live · free @sunaiva/gate

Sunaiva Gate

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

Live
@sunaiva/validation-gate-mcp

Triple-Gate Validation MCP

Patent-pending Triple-Gate runtime validation — USPTO 64/006,491 commercial embodiment.

AND-logic enforcement · cryptographic proof · MCP tool-callable

Live · PyPI
sunaiva-memory

Sunaiva Memory

Persistent semantic recall across AI agent sessions.

MCP-native · model-agnostic · team graphs · multi-project

Live · PyPI
sunaiva-ship-confidence

Sunaiva Ship Confidence

14-layer pre-deploy audit with HMAC-signed verdict.

Exit code 0/1/2 CI gates · Article 12 dossier export · audit hash chain


Why it holds

Documentation is a suggestion. A gate is a decision.

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.


        
01 Deterministic, not advisory

Rules evaluate locally in under a millisecond. Same input, same verdict, every time — no model in the hot path for the free tier.

02 Fail-safe by default

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.

03 MCP-native

Speaks the Model Context Protocol directly. Tool-callable by the generating model — validation invoked during generation, not bolted on after.

04 Cryptographically bound

Triple-Gate and Ship Confidence emit signed verdicts: gate signatures, output hash and timestamp bound into a composite proof.


Pricing

Free to start. $49/mo to go Pro.

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).

Honest framing Sunaiva developer tiers are priced low because they're freemium-funnel products — we want you trying the architecture. Enterprise validation, compliance and audit work lives at sunaiva.ai with appropriate enterprise pricing.