OpenRouter

OpenRouter is a unified gateway in front of hundreds of models from many vendors. The model slug is vendor-prefixed, e.g. anthropic/claude-sonnet-4, openai/gpt-4o, or google/gemini-2.5-pro.

chat-openrouter speaks both OpenAI-compatible wires and lets you pick one in the builder:

• Responses API (Beta) — the default. Wraps chat-responses and returns a ResponsesClient.
• Chat Completions — opt in with .with_completions(). Wraps chat-completions and returns a ChatCompletionsClient.

Install

[dependencies]
chat-rs = { version = "0.5.1", features = ["openrouter"] }

Or depend on the standalone crate: chat-openrouter = "0.1.0".

Usage

use chat_openrouter::OpenRouterBuilder;

// Default: Responses API.
let client = OpenRouterBuilder::new()
    .with_model("anthropic/claude-sonnet-4")
    .build();

// Opt in to Chat Completions instead.
let client = OpenRouterBuilder::new()
    .with_completions()
    .with_model("openai/gpt-4o")
    .build();

Set OPENROUTER_API_KEY in your environment or call .with_api_key().

Notes

• Stateless. The OpenRouter Responses API persists no server-side state and has no previous_response_id round-trip, so the builder always sends the full conversation each turn.
• Reasoning. .with_reasoning_effort("high") is available on the Responses wire for reasoning-capable models (Chat Completions has no such field).
• No WebSocket. OpenRouter is HTTP + SSE only; the builder stays generic over Transport, so a custom transport can still be supplied via .with_transport(...).