Let AI code. Stay in control.

Luthien is a proxy that sits between Claude Code (client) and the Claude API backend. It logs every request and response and enables you to set arbitrary rules/policies that can block dangerous operations, confirm the output is what you asked for, adheres to your claude.md and doesn't contain suspicious stuff, all without changing your dev setup.

ux
Before
Before: Claude Code runs pip install despite CLAUDE.md rules
After
After: pip install is blocked by Luthien, Claude retries with uv add
architecture
🧑‍💻 You ⇄ 💻 Claude Code ⇄ Luthien ⇄ ☁️ Anthropic API
                                    |
                             logs every request and response.   ~50ms
                             you can configure arbitrary rules/policies to
                             modify or block certain responses or requests:
                                    |
                                    |-- did it do what I asked?
                                    |-- did it follow CLAUDE.md?
                                    +-- did it do something suspicious?
                                    |
                                    +--> ☁️ can call another LLM (e.g. Haiku)
                                         to evaluate your rules/policies
See GitHub to create your own rules/policies

Open source. Early stage.
Book a call, and we'll help you set this up for your use cases.

Book a Call

Apache 2.0 · design partners wanted · contact@luthienresearch.org