One line change. Every LLM call protected.
CollieAi is an AI firewall that sits between your app and any LLM provider. Swap the base URL - every request and response is inspected, filtered, and logged by the content policies you define. No SDK, no library, no code changes.
client = openai.OpenAI(
base_url="https://api.collieai.io/v1", # ← that's it
api_key="your-openai-key"
)🛡️ Prompt injection detection · 🔒 PII & financial data masking · ⛔ Jailbreak prevention · 💬 Topic restriction policies · 🔍 Custom word filtering · 🔗 URL & link filtering · 🔑 Secret & API key detection · 🧠 Hallucination & quality checks · 📊 Full request/response logging
Unlike SDK-based guardrails, CollieAi works at the network level c any language, any framework, any LLM provider. ~12ms added latency per rule.
→ Get started free · Documentation · Quick Start · llm-firewall repo
AI firewall · LLM security · content guardrails · prompt injection protection · AI gateway · PII detection · content moderation · jailbreak detection · content filtering · LLM observability · AI safety · responsible AI · machine learning security · hallucination detection · word filtering · topic restriction · LLM proxy