CollieAi is a generative AI security and AI security posture management platform - an AI firewall, AI guardrails platform and LLM security gateway that protects production AI applications, chatbots, RAG pipelines, AI agents and agentic AI.
Provider-agnostic, it works with any LLM or AI system - OpenAI, Anthropic Claude, Google Gemini, Azure OpenAI, AWS Bedrock, deepseek and self-hosted models - and secures agentic AI / MCP (Model Context Protocol) and tool-calling workflows.
As a real-time AI security proxy it applies runtime guardrails to every request and response: prompt injection and indirect prompt injection protection, jailbreak detection, PII and personal information detection, redaction and masking, sensitive data and secrets / API-key detection, data loss prevention (DLP), content moderation, malicious URL filtering, and flexible custom word filtering with configurable dictionaries (brand safety, competitor names, sensitive terms). It runs in three modes - drop-in proxy, real-time streaming (SSE), and asynchronous (webhook) - with bidirectional input and output filtering, audit logging, observability, alerts and SIEM integration, delivering OWASP LLM Top 10 compliance and supporting GDPR, PCI DSS, HIPAA and SOC 2.
Deploy as a managed cloud service or fully self-hosted for complete data control.
