Injection firewall
A regex fast path sheds high-volume attacks at zero ML cost; an embedding-based semantic classifier judges the subtle ones. Three operating modes tune recall against false positives.
AI agent security · ERC-8004 native
Every agent makes promises.
Every agent makes promises.
GuardianAI proves them.
A dual-layer control plane for autonomous agents — millisecond prompt-injection firewalls off-chain, verifiable identity and tamper-evident evidence anchoring on-chain.
99.0% AdvBench strict|41.88 ms p95 under attack
guardian://gate/live-demo
awaiting input
Browser demo running a handful of public rules. The production gateway runs 40+ real-time controls server-side, with an embedding classifier behind the regex fast path.
Prompt injection
Base64 payload
Braille steganography
Morse encoding
ROT13 wrapper
Homoglyph swap
Hex payload
System-prompt leak
PII egress
XSS in response
99.0%
AdvBench block rate · strict
benchmark run · aug 2026
41.88 ms
p95 latency while blocking
chaos report · aug 2026
40+
real-time security controls
in one gateway pass
10/10
OWASP LLM Top 10 (2025)
risks covered · test snapshot
Threat model
Four layers between your agent and a bad day.
Each layer is independently tested and artifact-documented. Together they close the attack classes that ship in the wild today.
De-obfuscation engine
Morse, Braille steganography, Base64, hex, binary, ROT13, homoglyphs — normalized before classification so payloads can't hide behind an encoding layer.
Output guardrails
Responses scanned before egress: PII redacted, XSS/SQL/shell fragments blocked, system-prompt leakage caught, optional watermarking applied downstream.
Fail-closed operations
Rate limiting that closes instead of opens under failure, chaos-tested degradation, and SLO checks that run in CI — when parts die, the gateway fails safe.
One control plane
Security off-chain. Trust on-chain.
The execution layer decides in milliseconds whether traffic is safe. The trust layer makes an agent's identity and track record independently verifiable.
off-chain // execution security
- Fast-path screen — heuristics shed obvious attacks before any model inference
- Semantic firewall — embedding classifier, strict / balanced / permissive modes
- De-obfuscation pass — encodings decoded before judgment
- Response guardrails — redaction, payload blocking, watermarking
- Metering & limits — per-tenant tiers enforced fail-closed
on-chain // trust suite
- ERC-8004 identity — registered in the Trustless Agents ecosystem; ownership transfers with the agent
- Soulbound passport — ERC-5192, trust tiers UNVERIFIED → DIAMOND, score 0–10000
- Evidence anchoring — Merkle roots over security-event logs (simulated by default; testnet opt-in)
- Insurance certificates — signed evidence packages anchored to GuardianInsuranceLedger
- Interlock registry — explicit agent-to-agent authorization
Evidence
Numbers with receipts. Including the ugly ones.
Every figure on this site cites its dated repository artifact — and the proof page publishes full benchmark tables, not just highlights.
Benchmarks →
3,211 prompts across 8 public datasets, August 2026. AdvBench 99.0% strict, JBB PAIR+GCG 98.7% — and Do-Not-Answer's 57.1% published too, because selective reporting is how trust dies.
Audit status →
External pentest closed March 2026, zero open critical/high findings. Internal off-chain and contract audits remediated July–August 2026. No current third-party audit — stated plainly until Q3's external review lands.
Chains & identity →
Nine-contract suite on Base Sepolia and Monad testnet. Anchoring runs simulated-by-default. No mainnet claims anywhere on this site — check us.
Pricing
Hosted tiers are coming.
Hosted tiers are coming.
The engine is free now.
Managed platform
The full platform runs as a managed cloud service — no setup, no maintenance. Onboarding opens soon.
Hosted API tiers — coming soon
A managed gateway with usage-based tiers is in the works. When the meters run live, prices get published here — from the metering service, not from marketing.