Docs

Integrations

Any client that speaks the Anthropic Messages, OpenAI Chat Completions or OpenAI Responses API works. Change the base URL, set the key, and use a tier as the model.

Every example reads the key from ROUTERLANE_API_KEY. Leave reasoning and thinking settings at their defaults: the router replaces them per task.

Claude Code

Set these in your shell, then run claude. The [1m] suffix tells Claude Code the context window is 1M tokens; the router ignores it.

shell

export ANTHROPIC_BASE_URL=https://api.routerlane.com
export ANTHROPIC_AUTH_TOKEN=$ROUTERLANE_API_KEY
export ANTHROPIC_MODEL='smart[1m]'
export ANTHROPIC_DEFAULT_OPUS_MODEL='smart[1m]'
export ANTHROPIC_DEFAULT_SONNET_MODEL='normal[1m]'
export ANTHROPIC_DEFAULT_HAIKU_MODEL=fast
claude

To keep the setup with Claude Code instead, put the same variables under "env" in ~/.claude/settings.json, and keep ANTHROPIC_AUTH_TOKEN in your shell so the key stays out of the file.

~/.claude/settings.json

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://api.routerlane.com",
    "ANTHROPIC_MODEL": "smart[1m]",
    "ANTHROPIC_DEFAULT_OPUS_MODEL": "smart[1m]",
    "ANTHROPIC_DEFAULT_SONNET_MODEL": "normal[1m]",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "fast"
  }
}

Claude Code's background calls use the Haiku slot, so they run on Fast. To run a session as Smart+, start it with ANTHROPIC_MODEL=smart-plus claude.

Codex

Codex uses the Responses API. Add a provider to its config, export ROUTERLANE_API_KEY in the shell that runs Codex, and pick a tier with model or codex -m smart.

~/.codex/config.toml

model = "normal"
model_provider = "routerlane"

[model_providers.routerlane]
name = "RouterLane"
base_url = "https://api.routerlane.com/v1"
env_key = "ROUTERLANE_API_KEY"
wire_api = "responses"

For scripted runs, close stdin: codex exec "run the tests and fix what fails" < /dev/null.

pi

Add RouterLane as a provider in pi's models file, then choose Fast, Normal, Smart or Smart+ in pi's model picker. pi expands $ROUTERLANE_API_KEY from the environment, so the key stays out of the file.

~/.pi/agent/models.json

{
  "providers": {
    "routerlane": {
      "baseUrl": "https://api.routerlane.com",
      "api": "anthropic-messages",
      "apiKey": "$ROUTERLANE_API_KEY",
      "models": [
        {"id": "fast", "name": "Fast", "reasoning": false, "input": ["text", "image"], "contextWindow": 1048576, "maxTokens": 131072},
        {"id": "normal", "name": "Normal", "reasoning": false, "input": ["text", "image"], "contextWindow": 1048576, "maxTokens": 131072},
        {"id": "smart", "name": "Smart", "reasoning": false, "input": ["text", "image"], "contextWindow": 1048576, "maxTokens": 131072},
        {"id": "smart-plus", "name": "Smart+", "reasoning": false, "input": ["text", "image"], "contextWindow": 1048576, "maxTokens": 131072}
      ]
    }
  }
}

OpenAI SDK

Point the client at https://api.routerlane.com/v1. Chat Completions and Responses both work.

openai_example.py

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.routerlane.com/v1",
    api_key=os.environ["ROUTERLANE_API_KEY"],
)

# Read the routing decision from the headers
raw = client.chat.completions.with_raw_response.create(
    model="normal",
    messages=[{"role": "user", "content": "Write a unit test for slugify()."}],
)
print(raw.headers.get("x-router-model"), raw.headers.get("x-router-effort"))
completion = raw.parse()
print(completion.choices[0].message.content)

openai_example.ts

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.routerlane.com/v1",
  apiKey: process.env.ROUTERLANE_API_KEY,
});

// Read the routing decision from the headers
const { data, response } = await client.chat.completions
  .create({
    model: "normal",
    messages: [{ role: "user", content: "Write a unit test for slugify()." }],
  })
  .withResponse();
console.log(response.headers.get("x-router-model"), response.headers.get("x-router-effort"));
console.log(data.choices[0].message.content);

Anthropic SDK

Point the client at https://api.routerlane.com, without /v1.

anthropic_example.py

import os
import anthropic

client = anthropic.Anthropic(
    base_url="https://api.routerlane.com",
    api_key=os.environ["ROUTERLANE_API_KEY"],
)

with client.messages.stream(
    model="smart",
    max_tokens=8192,
    messages=[{"role": "user", "content": "Design a retry policy for a flaky payment API."}],
) as stream:
    for text in stream.text_stream:
        print(text, end="", flush=True)

anthropic_example.ts

import Anthropic from "@anthropic-ai/sdk";

const client = new Anthropic({
  baseURL: "https://api.routerlane.com",
  apiKey: process.env.ROUTERLANE_API_KEY,
});

const stream = client.messages.stream({
  model: "smart",
  max_tokens: 8192,
  messages: [{ role: "user", content: "Design a retry policy for a flaky payment API." }],
});
stream.on("text", (text) => process.stdout.write(text));
await stream.finalMessage();

Vercel AI SDK

Use the OpenAI-compatible provider.

ai_sdk_example.ts

import { createOpenAICompatible } from "@ai-sdk/openai-compatible";
import { generateText } from "ai";

const routerlane = createOpenAICompatible({
  name: "routerlane",
  baseURL: "https://api.routerlane.com/v1",
  apiKey: process.env.ROUTERLANE_API_KEY,
});

const { text } = await generateText({
  model: routerlane("normal"),
  prompt: "Write a commit message for: rename getUser to fetchUser.",
});
console.log(text);

LangChain

Use ChatOpenAI with RouterLane's base URL.

langchain_example.py

import os
from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    model="smart",
    base_url="https://api.routerlane.com/v1",
    api_key=os.environ["ROUTERLANE_API_KEY"],
)
print(llm.invoke("Plan a zero-downtime schema migration.").content)

curl

shell

curl -i https://api.routerlane.com/v1/messages \
  -H "x-api-key: $ROUTERLANE_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{"model": "fast", "max_tokens": 1024, "messages": [{"role": "user", "content": "hello"}]}'

-i prints the headers, so you can see the tier, model and effort the router chose.