CapabilitiesHow It WorksTrustIntegrationsSecurityPricing
Runtime Security Layer for AI Agents

Control what your AI agentscan detect

Zencorix sits inline between prompt, tool call, and execution so teams can secure AI agents operating in wallets and trading flows, internal tools, and other high-trust environments.

InlineDetect prompt injection, jailbreak,and poisoning signals at ingress
RuntimeEnforce tenant policy with allow,block, or freeze before tool execution
ProofsSigned receipts and immutableaudit trail for every security decision
TransactionDetect drain patterns and preventhigh-blast-radius actions in real time
Execution RiskEnforce spend limits, allowlists and approval gatesfor DeFi AI agent wallets
Capabilities

Runtime
governance.

One product story, three clear buyers: wallet and trading agents, tool-using copilots, and security/audit teams.

Best wedge today is DeFi and wallet-agent protection, while the same runtime architecture fits AI systems outside crypto.

01

Prompt & Action Detection

Classify risky intent across user prompts and tool/action contexts before execution, including prompt injection and indirect injection.

Detectinline risk evaluation
Flow

Detect.Enforce.Prove.

Trust infrastructure
Global network sphere

Built first
for DeFi agents.

DeFi is the wedge because if an agent can approve spend, route swaps, or move assets, guardrails must run before the transaction is signed.

Runtimedefense

High-blast-radius wallet actions are checked in-path before execution.

Walletapproval and transfer risk controls
Sessionchained-abuse pattern analysis
enabled
Infinite Approval BlockingFlags and blocks open-ended spend approvals before signing.
enabled
Destination Risk HeuristicsFlags risky destination patterns and policy violations before transfers.
enabled
Approve-Then-Drain DetectionTracks session behavior to catch chained approval and transfer abuse.
enabled
Honeypot Token DefenseHighlights token behaviors that do not match normal wallet expectations.
enabled
Extreme Slippage GuardStops dangerous swap configurations before they land on-chain.
enabled
Multi-Token Sweep AlertRaises alerts when sessions resemble coordinated wallet-drain sequences.
Proof & Ops

Proof
& operating model.

Runtime-grade
Detection posture
Calibrated for high-trust agent workflows
Policy-verified
Enforcement posture
Deterministic controls before side effects
Hash-first
Sensitive-data-aware proofs
hashes + attestations, not raw prompt text
Centralized runtime control plane todayAsync proof commitmentReplayable verification for incident0G proof anchoring as trust layer
Integrations

Integrate
your stack.

Live integrations currently supported in production paths. We only list what is actually wired today.

Runtime
🐍
🐍 Python
Runtime
🟨
🟨 JavaScript
Framework
🦜
🦜 LangChain
Agent
🤖
🤖 AutoGPT
Platform
⚡
⚡ CustomGPTs
Framework
LlamaIndex logo
LlamaIndex
Roadmap
+
Coming SoonOther integrations are coming soon.
PythonSDK and API integration path
JavaScriptSDK and API integration path
Agent FrameworksLangChain, AutoGPT, LlamaIndex
Security

Governed,
not blind.

Runtime security changes execution outcomes. Move from after-the-fact monitoring to policy enforcement before side effects.

Without Zencorix

  • Prompt injection reaches live tools and downstream systems
  • Agents execute risky approvals, transfers, or destructive actions
  • Sensitive context leaks into prompts, logs, or memory
  • Multi-step abuse chains are detected after execution
  • Security teams lack replayable records for disputes
  • High-risk workflows rely on scattered app-side checks

With Zencorix

  • Inline validation before prompts or tool calls execute
  • Risk scoring across patterns, entropy, ML, and context signals
  • Deterministic runtime enforcement for high-risk actions
  • Session-aware analysis for chained behavior
  • Hash-first proof flow keeps sensitive text off-chain
  • Replayable audit trail with async proof anchoring on 0G
Detection coverage
Real-time attack surface protectionRuntime policies evaluate threats before agents execute sensitive actions.
Prompt InjectionData PoisoningModel PoisoningMCP/Tool Poisoning

Prompt Injection Detection

Detect direct and indirect prompt injection attempts before tool execution.

Data Poisoning Signals

Flag suspicious retrieval context and manipulated knowledge inputs.

Model Poisoning Indicators

Surface abnormal behavior patterns that suggest compromised model outputs.

Tool & MCP Poisoning

Detect malicious tool payloads, unsafe function calls, and scope violations.

Developer SDK

Secure your agents.
Operate with confidence.

One runtime layer for builders and security teams: detect inline, enforce before side effects, and preserve verifiable evidence without slowing execution.

Runtime SDK + API

Drop-in validate/check/execute flow for tool-calling agents.

Tenant policy control

Versioned rules, scoped overrides, and approval pathways.

Proof-first operations

Signed receipts and replayable audit evidence per decision.

Security team workflows

Incident review, dispute lookup, and governance-ready history.

Pricing

Plans for
secure scale.

Organic whale
01

Starter

For early teams

$499/month

billed annually

  • 100,000 runtime checks / month
  • Up to 3 protected agents
  • Validator + runtime enforcement
  • Metadata audit logs
  • Email support
  • Overage: $8 per 10,000 checks
Most Popular
02

Growth

Most Popular

$1999/month

billed annually

  • 750,000 runtime checks / month
  • Up to 12 protected agents
  • Override workflows
  • Proof trail (Chain + Storage)
  • Incident workflow tooling
  • Priority support
  • Overage: $6 per 10,000 checks
03

Enterprise

For regulated teams

Custom
  • 3,000,000 runtime checks / month
  • Up to 50 protected agents (custom above)
  • SSO + RBAC
  • Custom retention posture
  • DA batching and compliance exports
  • SLA and dedicated success manager
  • Advanced governance controls
  • Overage: $4 per 10,000 checks

Add-ons

  • •Extra protected agents: $20/agent/mo (Starter), $15/agent/mo (Growth), custom (Enterprise)
  • •High-assurance recheck (Compute path): $0.003 per uncertain-case recheck
  • •Compliance export pack: $500/mo
  • •24/7 premium support: $1,000/mo
  • •One-time onboarding: $2,500 (Starter/Growth), $10,000 (Enterprise)

Pilot Offer

  • •30-day pilot: $1,500 fixed
  • •Includes up to 300,000 checks and up to 10 agents
  • •Proof trail enabled and runtime enforcement integrated
  • •Security report delivered at pilot end

Start securing
live agent workflows.

Start with hands-on validation, or walk buyers through proof verification and runtime architecture for high-trust environments.

Built for high-trust enterprise AI workflows