AUTONOMOUS AIAS A SERVICE
Web3-native agents. Free and open.
Wallet is the gate. Skill-Guard scans every prompt before it ships.
// Data Sources
POWERED BY VERIFIED SOURCES
// Live Telemetry
Real-time $SCTR network stats
// The Stack
THE $SCTR ECOSYSTEM
Infrastructure for autonomous AI agents. Priced in crypto, verified on-chain, owned by you.
Deploy autonomous agents
Wallet-bound agents that reason, plan, execute across multi-step tasks.
Zero cost, unlimited
No subscriptions. No credit cards. No metering. Whole network is free to use.
Your skills, your stack
Define custom skill.md files. Compose prompts and tools without vendor lock-in.
Wallet is the API key
Deterministic keys derived from your wallet. No servers to compromise.
Real-time SSE delivery
Token-by-token responses over server-sent events. Built for chat UIs.
Production-grade perf
25+ models routed through an aggregation layer tuned for agent workloads.
// Skill Guard
SCAN BEFORE YOU SHIP
Every skill pushed to the $SCTR registry passes through six detectors. Prompt injection, credential leaks, RCE payloads — caught at write-time, not after a wallet drains.
Credential leak
Match 160+ secret patterns — API keys, private keys, JWTs, mnemonics — before a skill ships to the registry.
Prompt injection
75+ jailbreak heuristics including obfuscation, role-swap, and indirect injection through tool outputs.
Malicious command
112+ patterns for RCE, reverse shells, post-exploit cleanup, and dangerous shell idioms inside skill bodies.
Data exfiltration
Catch path traversal, env harvesting, DNS tunneling, and outbound webhook drops in tool definitions.
Permission abuse
Diff declared scopes vs. what the skill actually touches. Flag privilege creep before runtime.
URL & domain risk
Cross-check every URL against phishing feeds, typosquats, and known-malicious infra before forwarding.
// Three-step pipeline
Write → Scan → Publish
Prompt + tool spec, plain markdown.
Six detectors run in parallel, CWE-mapped.
Wallet-signed, on-chain manifest.
// Flow
THREE STEPS.
ZERO FRICTION.
From wallet connect to first agent call in under sixty seconds.
CONNECT WALLET
Sign in with any EVM wallet. No email. No password. Your address is your identity.
$ scrutor auth✓ Wallet connected✓ API key: sctr_live_••••1a2b
PICK A SKILL
Browse the on-chain skill.md registry. Compose prompts + tools. Deploy in one call.
POST /v1/agents→ skill: defi-snipe→ network: base✓ ready (12ms)
STREAM RESPONSES
Free SSE streaming. Zero cost. Nothing to cancel — nothing is subscribed.
GET /v1/stream← token.delta …← token.delta …✓ complete
// Token Utility
THE $SCTR TOKEN
$SCTR powers the Scrutor network. Fair-launched on Base.
No allocations. No vesting. No pre-sale.
Contributor rewards
Skill authors and model operators earn $SCTR from weekly emissions tied to usage.
Skill staking
Lock $SCTR to publish skills into the on-chain registry and earn attribution.
Governance
Route new models, vote on skill moderation, steer protocol upgrades.
// Ship Now
AI AGENTS.ON-CHAIN.
No signup. No API keys. Just a wallet.
One call away from production.
// FAQ
FREQUENTLY ASKED
Scrutor is a Web3-native agent infrastructure. Deploy autonomous AI agents that reason, plan, and execute multi-step tasks — free and open for anyone with a wallet. No servers to manage, no API keys to rotate, no subscriptions to cancel.
You don't. Your wallet is the API key. When you connect, a deterministic key is derived from your address (sctr_live_…). Same wallet, same key, always. Nothing is stored server-side.
$SCTR is the utility token. Publish custom skills into the on-chain registry, stake to back reputation, and vote on protocol upgrades. The network itself is free — $SCTR is for governance and contributors, not gating access.
Yes. Every skill that lands in the $SCTR registry is run through Skill-Guard — six detectors covering credential leaks (160+ patterns), prompt injection (75+ heuristics), malicious commands (112+ patterns for RCE / reverse shells), data exfiltration, permission abuse, and URL phishing. Findings are CWE-mapped and the whole sweep finishes in under 50ms.
CRITICAL severity blocks publish. HIGH and MEDIUM emit a warning with line-level annotations — you decide whether to ship. Detector rules are open and versioned; you can subscribe to the changelog or pin a ruleset hash for reproducible scans.
Base is primary. Ethereum mainnet is supported for identity and governance. More L2s coming after v2.
Yes. Write a skill.md, point it at your own model endpoint, and publish to the registry. Scrutor routes + bills, you serve. Enterprise tier ships with fully self-hosted runtime.
CLI, SDK, and the skill.md spec are MIT-licensed. Core routing layer is source-available for audit. The smart contracts are verified on BaseScan.