LocalRouter is a local, privacy-first API gateway that runs on your machine and provides a single OpenAI-compatible endpoint at localhost:3625. It routes requests across 19+ LLM providers (OpenAI, Anthropic, Gemini, Ollama, LM Studio, and more), proxies MCP servers through a unified gateway, and supports configurable model selection using decision providers and exact client mode rules.

LocalRouter has zero telemetry, bundles its assets, and stores secrets in your OS keychain. Requests go only to the providers you configure; choose local providers to keep inference on your machine.

Clients are API consumers that authenticate with LocalRouter via API keys (lr-* prefix) or OAuth tokens. Each client references a Strategy that defines which models, providers, and MCP servers it can access, along with rate limits.

Providers are upstream LLM services (cloud or local) that LocalRouter routes requests to.

MCP Servers are tool providers connected via STDIO, SSE, or Streamable HTTP transports — LocalRouter merges them behind a single endpoint with automatic tool namespacing (server__tool).

Skills are curated multi-step workflows exposed as MCP tools that compose multiple tool calls into a single action.

LocalRouter runs an HTTP server on port 3625 that serves two gateways at the same root path:

OpenAI-compatible gateway. Exposes standard endpoints (/chat/completions, /models, /embeddings, /systemone, etc.) that any OpenAI-compatible client can use. The routing engine authenticates the client, checks rate limits, selects a model (via auto-routing or explicit model ID), and dispatches to the appropriate provider.

MCP gateway. Manages connections to upstream MCP servers, namespaces their tools, and proxies JSON-RPC calls. Clients connect once and gain access to all configured MCP servers.

Monitoring. Collects per-request metrics (latency, tokens, cost) viewable from the dashboard.

The UI is a native desktop window for managing configuration, while API consumers interact through the HTTP server.