LunaRoute

LunaRoute is a high-performance local proxy designed to enhance the functionality and transparency of AI coding assistants like Claude Code, OpenAI Codex CLI, and OpenCode. It provides complete visibility into every interaction with AI models, making it an essential tool for developers working with AI coding assistants.

LunaRoute supports both OpenAI and Anthropic formats simultaneously with minimal latency. It offers comprehensive session recording, detailed session statistics, and automatic detection and redaction of sensitive information. Additionally, it provides Prometheus metrics for monitoring and a built-in web interface for browsing sessions.

Benefits

LunaRoute offers several key benefits for developers using AI coding assistants:

• Complete Visibility: LunaRoute provides complete visibility into every interaction with AI models, allowing developers to track and analyze every request and response.
• Comprehensive Session Recording: LunaRoute records every interaction with dual storage options, including an SQLite database for fast queries and analytics and JSONL logs for human-readable, full request/response data.
• Detailed Session Statistics: LunaRoute provides detailed breakdowns of sessions, including the number of requests, token usage, tool performance, response times, and overhead.
• PII Detection & Redaction: LunaRoute automatically detects and redacts sensitive information such as emails, SSNs, credit card numbers, and phone numbers before it hits disk.
• Prometheus Metrics: LunaRoute provides 24 different metric types at/metrics, which are perfect for Grafana dashboards. These metrics include request rates, latency histograms, token usage, tool call statistics, and streaming performance.
• Web UI: A built-in web interface is available for browsing sessions, featuring a dashboard with filtering and search, session details with timeline view, raw JSON inspection, and token usage and performance analytics.

Use Cases

LunaRoute can be used in various scenarios to enhance the functionality and transparency of AI interactions:

• Debugging Expensive Conversations: LunaRoute can help identify why an AI session cost more than expected by providing detailed session statistics.
• Identifying Performance Bottlenecks: LunaRoute can reveal performance bottlenecks in AI interactions, allowing developers to optimize their AI assistants.
• Team Collaboration: LunaRoute allows teams to share one proxy without sharing API keys, making it easier for teams to collaborate on AI projects.
• Privacy & Compliance: LunaRoute enables automatic PII redaction, ensuring that sensitive data is not stored in session logs.

Installation

LunaRoute can be installed in two ways:

• Option 1: Download Binary (Recommended): Download the latest release for your platform, extract and run the binary, and optionally add it to your PATH for global access. Start using it immediately with the commandeval $(lunaroute-server env).
• Option 2: Build from Source: Clone the repository, navigate to the directory, build the project withcargo build --release --package lunaroute-server, and start using it with the commandeval $(./target/release/lunaroute-server env).

Usage

The fastest way to get started with LunaRoute is with the one-command setup:

# One command does everything!eval $(lunaroute-server env)# Now use your AI tools - they're automatically configured# Both Claude Code and Codex CLI work immediately

For more advanced users, LunaRoute can be configured manually by starting the server in one terminal and configuring the shell in another terminal.

Advanced Configuration

LunaRoute offers several advanced configuration options, including quick setup with theenvcommand, environment variables for controlling behavior without config files, and connection pooling for tuning HTTP client performance.

Performance

LunaRoute is designed for high performance with minimal latency and overhead. In passthrough mode, it adds only 0.1-0.2ms of latency and uses less than 1% CPU idle and less than 5% CPU at 100 RPS. With session recording, it adds 0.5-1ms of latency and uses batched writes every 100ms.

Architecture

LunaRoute is built as a modular Rust workspace with several key crates, includinglunaroute-corefor types and traits,lunaroute-ingressfor HTTP endpoints,lunaroute-egressfor provider connectors with connection pooling,lunaroute-sessionfor recording and search,lunaroute-piifor PII detection/redaction,lunaroute-observabilityfor metrics and health, andlunaroute-serverfor the production binary.

Contributing

LunaRoute welcomes contributions from the community, including bug reports and fixes, new PII detectors, additional metrics, documentation improvements, and performance optimizations.

License

LunaRoute is licensed under the Apache License, Version 2.0.

Why "LunaRoute"?

Like the moon guides travelers at night, LunaRoute illuminates your AI interactions. Every request, every token, every decision - visible and trackable. Built with love for developers who want visibility, control, and performance.

About

LunaRoute is a high-performance local proxy for AI coding assistants like Claude Code, OpenAI Codex CLI, and OpenCode. It provides complete visibility into every LLM interaction with zero-overhead passthrough, comprehensive session recording, and powerful debugging capabilities.

Give your AI coding assistant the visibility it deserves.lunaroute-server