Team Bring-Your-Own Guardrails
Let developers register a Generic Guardrail API guardrail for their team on the LiteLLM AI Gateway, then approve or reject it in the Admin UI.
Aim Security
Add Aim Security guardrails to the LiteLLM AI Gateway: create an Aim Guard, tune its detections, and set per-key policies by virtual key alias.
Airia
Block prompt injection, jailbreaks and sensitive data with Airia on the LiteLLM AI Gateway, allowing, blocking or redacting prompts and responses.
Akto
Add Akto runtime security to the LiteLLM AI Gateway: inspect requests, responses and MCP calls inline, then block or monitor without blocking.
Alice
Screen prompts and responses with Alice on the LiteLLM AI Gateway: allow, block or replace flagged text, with policies per named application.
Aporia
Detect PII in requests and profanity in responses with Aporia on the LiteLLM AI Gateway, using pre-call and post-call projects and per-key control.
Azure Content Safety Guardrail
Block jailbreaks and harmful text with Azure Content Safety on the LiteLLM AI Gateway, using Prompt Shield and Text Moderation, with cost tracking.
Bedrock Guardrails
Apply Amazon Bedrock Guardrails on the LiteLLM AI Gateway via the ApplyGuardrail API, with PII masking, contextual grounding and streaming checks.
CollieAi
Check chat completions against a CollieAi project policy on the LiteLLM AI Gateway before and after the model call, in observe or block mode.
Context Compression (Compresr)
Use Compresr on the LiteLLM AI Gateway to compress tool outputs, retrieved documents and RAG payloads before they reach the model, per key or call.
Conduct Guard
Send each prompt to your Conduct workspace from the LiteLLM AI Gateway, blocking on block or approval verdicts and passing warnings through.
CrowdStrike AIDR
Block prompt injection, PII and toxic content with CrowdStrike AIDR policies on the LiteLLM AI Gateway, with collector registration and setup.
Custom Code Guardrail
Write guardrail logic as sandboxed Python-like code on the LiteLLM AI Gateway, with built-in regex and JSON helpers to block or flag requests.
Custom Guardrail
Write a CustomGuardrail class for the LiteLLM AI Gateway: implement apply_guardrail to block or modify content and register it in config.yaml.
Decision Model Guardrail
Ask a decision model such as TypeSafe Jev yes/no questions about each request on the LiteLLM AI Gateway, and block or log above a threshold.
DynamoAI Guardrails
Add DynamoAI content moderation and policy enforcement to LLM inputs and outputs on the LiteLLM AI Gateway, with policy IDs and a custom API base.
EnkryptAI Guardrails
Add EnkryptAI content moderation and safety checks to LLM inputs and outputs on the LiteLLM AI Gateway, with custom policies and observability.
Gray Swan Cygnal Guardrail
Monitor or block policy violations, jailbreaks and indirect prompt injection with Gray Swan Cygnal on the LiteLLM AI Gateway, including Claude Code traffic.
Guardrails AI
Run Guardrails AI guards on LLM calls through the LiteLLM AI Gateway: point config.yaml at your Guardrails AI server and control guards per key.
HiddenLayer Guardrails
Send every request and response on the LiteLLM AI Gateway to HiddenLayer to block or redact unsafe content, with SaaS or self-hosted credentials.
IBM Guardrails
Detect jailbreaks, PII and hate speech with IBM FMS Guardrails on the LiteLLM AI Gateway, calling a detector server or the orchestrator from config.yaml.
Javelin Guardrails
Add Javelin prompt injection detection, trust and safety checks and language detection to the LiteLLM AI Gateway as pre_call guardrails in YAML.
Lakera AI
Block prompt injection and mask PII with Lakera Guard on the LiteLLM AI Gateway: config.yaml setup plus pre_call, during_call and post_call modes.
Lasso Security
Block prompt injection and mask PII with Lasso Security on the LiteLLM AI Gateway, scanning inputs and outputs with user and conversation tracking.
LLM-as-a-Judge
Score each request or response 0-100 against weighted criteria with a judge model on the LiteLLM AI Gateway, and block or log calls below a threshold.
LLM Shield Proxy
Swap personal data for stand-in values before it reaches the model and restore it in the reply with LLM Shield Proxy on the LiteLLM AI Gateway.
Microsoft Agent 365 Guardrail
Send each MCP tool call from the LiteLLM AI Gateway to Microsoft Agent 365, so Microsoft Defender allows or blocks it and records it under the user.
Microsoft Purview Guardrail
Enforce Microsoft Purview DLP policies on the LiteLLM AI Gateway through the Microsoft Graph processContent API, with Entra app setup and user IDs.
Google Cloud Model Armor
Apply Google Cloud Model Armor templates on the LiteLLM AI Gateway to sanitize or block prompts and responses, including document and file scanning.
Noma Security
Add Noma Security moderation and safety guardrails to the LiteLLM AI Gateway with noma_v2, including MCP call modes and a monitor-only mode.
Onyx Security
Add Onyx Security AI Guard policies to the LiteLLM AI Gateway: create a policy, copy its API key and run it before, during or after the call.
OpenAI Moderation
Block hate speech, harassment, self-harm, sexual content and violence with the OpenAI Moderation API on the LiteLLM AI Gateway, with streaming.
Ovalix
Send prompts, responses, files and tool calls from the LiteLLM AI Gateway to the Ovalix Tracker, which blocks or anonymizes content by policy.
Pangea
Block prompt injection, PII and toxic content with Pangea AI Guard recipes on the LiteLLM AI Gateway, with separate input and output recipes.
PANW Prisma AIRS
Scan prompts and responses with Palo Alto Networks Prisma AIRS on the LiteLLM AI Gateway, with security profiles, content masking and fail-open.
PII, PHI Masking - Presidio
Mask PII and PHI in LLM requests and responses with Microsoft Presidio on the LiteLLM AI Gateway, with entity types and confidence thresholds.
Pillar Security
Block prompt injection, jailbreaks and PII leaks with Pillar Security on the LiteLLM AI Gateway via the Generic Guardrail API, on prompts and responses.
In-memory Prompt Injection Detection
Detect prompt injection on the LiteLLM AI Gateway with in-memory similarity checks or an LLM API check, across chat, messages and responses.
PromptGuard
Detect prompt injection, redact PII and filter topics with PromptGuard on the LiteLLM AI Gateway, with fail-open mode and always-on protection.
Qostodian Nexus by Qohash
Block or redact sensitive data in prompts and model output with Qohash Qostodian Nexus on the LiteLLM AI Gateway, run as a container you deploy.
Qualifire
Detect prompt injection, hallucinations, PII and harmful content with Qualifire on the LiteLLM AI Gateway, with evaluations and custom assertions.
Reco
Send prompts from the LiteLLM AI Gateway to Reco for inspection before they reach the model, so Reco can block or flag risky requests in pre_call mode.
RepelloAI Argus
Scan prompts and responses with RepelloAI Argus on the LiteLLM AI Gateway against the policies you set per asset in the Repello dashboard, with streaming.
Rubrik Guardrail
Screen prompts and responses with Rubrik moderation on the LiteLLM AI Gateway in OpenAI and Anthropic formats, and batch-log every LLM request.
Secret Detection/Redaction (Enterprise-only)
Redact API keys and secrets from requests on the LiteLLM AI Gateway before they reach an LLM, turned on per key or team, with detector control.
Sensitive Data Routing (Built-in Guardrail)
Detect sensitive data in a request on the LiteLLM AI Gateway and reroute it to an on-premise model instead of blocking it, with session stickiness.
Straiker
Block prompt injection, including payloads hidden in tool output, with Straiker on the LiteLLM AI Gateway, inspecting prompts and tool calls.
LiteLLM Tool Permission Guardrail
Control which tool calls a model may make on the LiteLLM AI Gateway with allow and deny rules for OpenAI tool_calls, Anthropic tool_use and MCP.
Relevance-Based Compaction (TypeSafe / Jev)
Blank out tool results that TypeSafe Jev scores below a relevance threshold on the LiteLLM AI Gateway, so stale context stops using input tokens.
Vigil Guard
Evaluate chat input and model output with Vigil Guard on the LiteLLM AI Gateway before results return to your app, with run-by-default and fail-open.
Wingback
Block prompt injection, jailbreaks and PII leaks with Wingback on the LiteLLM AI Gateway, inspecting prompts and responses to monitor or enforce.
XecGuard
Scan for prompt injection, harmful content and PII with CyCraft XecGuard on the LiteLLM AI Gateway, plus RAG context grounding and fail-open.
Zscaler AI Guard
Enforce Zscaler AI Guard policies on prompts and responses through the LiteLLM AI Gateway, with custom policy IDs per team or key and user details.