Quick Start - Gateway
Stand up the gateway, connect a provider, and make your first guarded LLM call in under five minutes.
DeepintShield
Real-time guardrails for prompts, agents, and tools - authorize every tool-call, cut LLM spend, and catch PII, jailbreaks, and hallucinations before they reach your users.
<5ms p50Guardrail latency
~2µsCached tool-call verdict
Up to 90%LLM cost saved
2500+ models24+ providers, one API
OWASPLLM + Agentic Top-10
MCPGateway & tool control
Agent IDIdentity & auth
Zero-configAgent discovery
Real-timeHallucination checks
Start here
Section titled “Start here”Performance & Cost
Five optimizations ship enabled (or one flag away): embedded runtime, speculative dispatch, async post-guards, per-category timeouts, semantic cache short-circuit.
AI Guardrails
Configure safety policies across all five evaluation stages - input, output, MCP action, RAG retrieval, and post-LLM.
SDK Integrations
Drop-in adapters for OpenAI, Anthropic, Bedrock, Gemini, LiteLLM, LangChain, and PydanticAI.
What makes DeepintShield different
Section titled “What makes DeepintShield different”Real-time, in-process
Guard evaluation runs in the gateway itself - no RPC hop, no sidecar. Embedded runtime is on by default in single-binary deployments.
Speculative dispatch
Fire the provider call in parallel with input guards. Allow-path latency becomes
max(guards, model) instead of guards + model.
Smart caching cuts spend
Semantic cache short-circuits the guard + provider pipeline on templated traffic. Up to 60% cost reduction on chatbot-style workloads.
Auditable by default
Every decision lands in the evidence store - sync, shadow, and async findings alike. Export-ready for SOC 2 / ISO 27001.
Get involved
Section titled “Get involved”GitHub
Star, file issues, or send PRs.
Discord
Real-time help, roadmap discussion, design reviews.