what's MoltWall?

MoltWall is a production-grade security firewall for AI agents and agentic workflows. It serves as middleware between agents and execution environments (wallets, APIs, MCP servers), intercepting tool calls to evaluate risks and enforce policies before action. The deterministic policy engine checks allowlists, limits, and domains via Redis-cached rules. A risk engine scores requests (0-1) across factors like provenance and intent, while guardrails scan for prompt injections, tool poisoning, credential leaks, and PII in nested args. Decisions—allow, deny, sandbox, require_confirmation—are returned in <10ms with explanations and immutable Supabase audit logs. The TypeScript SDK (@moltwall/sdk) enables easy integration with LangChain, AutoGPT, CrewAI, LangGraph, and custom frameworks via a single wall.check() call. Deployed on 5 global edge instances for low latency, it uses Upstash Redis for caching/rate-limiting and Supabase for logging/queries. Roadmap advances to developer platform (Python/Go SDKs, YAML policies, webhooks), enterprise RBAC/notifications, ecosystem integrations, and intelligence (telemetry, heatmaps).

MoltWall completed Phase 1, now in Phase 2 with a developer platform.