SECURE EVERYAGENT ACTIONBEFORE IT FIRES
MoltWall is a production-grade security firewall for AI agents. Every tool call evaluated, every threat blocked, every decision audited -in under 10ms.
Core Capabilities
EVERYTHING AN AGENT
FIREWALL NEEDS
Full-Stack Firewall
Every agent action passes through a deterministic policy engine before execution. Allow, deny, sandbox, or require confirmation.
Real-Time Monitoring
Every request is scored, every decision logged with full provenance. Complete audit trail across all agents and tools.
Threat Guardrails
Prompt injection, credential leaks, PII exposure, and tool poisoning detected and blocked before damage occurs.
Policy Engine
Define allowed tools, blocked actions, trusted domains, and spend limits. Redis-cached, sub-millisecond enforcement.
Risk Scoring
Multi-factor 0–1 risk score computed per-request. Source provenance, payload analysis, intent matching all factored in.
SDK & API
Drop-in TypeScript SDK. One function call integrates MoltWall into any MCP agent, LangGraph flow, or custom framework.
The Pipeline
HOW MoltWall
PROTECTS YOU
Agent Request
Your agent calls a tool. The MoltWall SDK intercepts the call before execution.
Policy Check
Tool allowlist, blocked actions, and spend limits are evaluated instantly from Redis cache.
Risk Score
Payload is scored across 8 weighted factors including source provenance and argument analysis.
Guardrail Scan
Prompt injection, credential patterns, and PII are scanned recursively across nested arguments.
Decision
Allow, Deny, Sandbox, or Require Confirmation -returned in <10ms with a full explanation.
Audit Log
Every decision persisted to Supabase with full trace. Query and export from the dashboard.
Quick Integration
ONE CALL TO
FIREWALL YOUR AGENT
Drop the MoltWall SDK into any TypeScript agent. Works with Claude MCP, LangChain, AutoGPT, CrewAI, and any custom framework. Zero config firewall in one call.
import { MoltWall } from "@moltwall/sdk";
const wall = new MoltWall({
apiKey: process.env.MOLTWALL_API_KEY,
baseUrl: "https://www.moltwall.xyz",
});
// Before every tool call:
const result = await wall.check({
action: "transfer_funds",
tool: "wallet",
args: { amount: 100, to: addr },
source: "user",
});
if (result.decision === "allow") {
await executeTool(result);
} else {
// denied, sandbox, or require_confirmation
handleBlocked(result);
}Start Today
YOUR AGENTS.
FIREWALLED.
Deploy MoltWall at www.moltwall.xyz in minutes. Open source. TypeScript-native. Production firewall from day one.