A drop-in proxy that sits between your application and any LLM. Sensitive data never reaches Claude, GPT-4o, or Gemini. Change one URL. Zero other code changes.
Jnvoy intercepts every LLM call before it leaves your system. Sensitive data is detected, replaced with tokens, and restored in the response. Nothing sensitive ever travels.
Normal LLM call with real user data
Presidio detects 20+ sensitive entity types locally
Replaced with reversible tokens before transmission
Claude or GPT-4o never sees raw sensitive data
Original values restored. Audit log written. Done.
Everything you need to deploy AI safely in a regulated environment. No configuration required. No vendor lock-in.
Enterprise-grade NLP-based PII detection. 20+ entity types. UK-specific recognisers for NHS numbers, NI numbers, UK postcodes, and sort codes.
Similar queries return cached responses instantly. Saves 20–40% of LLM costs automatically without any configuration.
Assigns the cheapest capable model to each query automatically. Simple questions go to Haiku. Complex ones go to Sonnet.
Every API call permanently logged to PostgreSQL. Never stores raw PII. Queryable by date, provider, and PII type.
One endpoint generates an FCA and GDPR-ready PDF proving zero sensitive data was transmitted. With your company name on it.
Each company gets complete data isolation. Their traffic, audit logs, and compliance reports are entirely separate.
Outage detector tracks provider health. If Claude goes down, traffic automatically reroutes to GPT-4o or Gemini.
Protects every message sent to any AI bot in your Slack workspace. Posts a compliance note showing what was detected.
One command starts everything. API plus PostgreSQL plus Redis. Self-hostable on your own infrastructure.
No SDK to install. No configuration to learn. No architectural changes. Point your existing LLM calls at Jnvoy and every request is automatically protected.
# Before Jnvoy import anthropic client = anthropic.Anthropic( api_key="sk-ant-..." ) response = client.messages.create( model="claude-haiku-4-5", messages=[{ "role": "user", "content": user_message # ⚠️ raw PII }] )
# After Jnvoy — one line changed import httpx response = httpx.post( "https://jnvoy.onrender.com/v1/proxy", json={ "model": "claude-haiku-4-5", "provider": "anthropic", "messages": [{ "role": "user", "content": user_message # ✓ PII redacted }] } )
When the FCA or ICO asks how you ensure sensitive data never reaches external AI models, Jnvoy generates the answer automatically. One endpoint. One PDF. Done.
"We built PII stripping internally and it took weeks. The problem is real. The larger enterprise partners always ask about it before signing."
— Co-founder, Series A fintech AI company · unsolicited feedback
Open source. Self-hostable. Works with every LLM.