Get started with BedrockRouter

Architecture

BedrockRouter is a thin, OpenAI-compatible proxy that is a centralized proxy. Requests authenticate with an API key, get translated into the Bedrock InvokeModel / Converse protocol, and stream straight back to your client.

Quickstart

Get a working request in under five minutes.

1. 1

   Create an account

   Sign up at BedrockRouter to track your deployments and integrations.

   Create account →
2. 2

   Deploy the CloudFormation stack

   Click Deploy, follow the AWS console wizard, and wait ~60 seconds for the stack to finish.

   Deploy stack →
3. 3

   Get your proxy URL and API key

   Once the stack finishes, your proxy URL appears on your dashboard. Create an API key from the API Keys page — no AWS console needed.

4. 4

   Use the BedrockRouter SDK

   Install the official SDK and start building in seconds:

   bash

   pip install bedrockrouter

   Then use it with your API key:

   python

   from bedrockrouter import BedrockOpenAI, Models
   
   client = BedrockOpenAI(api_key="YOUR_API_KEY")
   
   resp = client.chat.completions.create(
       model=Models.CLAUDE_SONNET_4_6,
       messages=[{"role": "user", "content": "Hello, Bedrock!"}],
   )
   print(resp.choices[0].message.content)

   See more SDK examples →

SDK examples

Use the official BedrockRouter SDK for the best experience with 100+ models and friendly aliases. Or use OpenAI-compatible and Anthropic-compatible SDKs by changing the base URL. Go native with boto3 if you prefer.

BedrockRouter Python SDK (Official)

The easiest way to use Bedrock models. Two compatible clients, 100+ models with friendly aliases, streaming support, and built-in cost tracking.

pip install bedrockrouter

from bedrockrouter import BedrockOpenAI, Models

client = BedrockOpenAI(
    api_key="YOUR_API_KEY",
)

response = client.chat.completions.create(
    model=Models.CLAUDE_SONNET_4_6,
    messages=[{"role": "user", "content": "Explain Bedrock in one sentence."}],
    max_tokens=1024,
)

print(response.choices[0].message.content)

from bedrockrouter import BedrockRouter, Models

client = BedrockRouter(
    api_key="YOUR_API_KEY",
)

message = client.messages.create(
    model=Models.CLAUDE_SONNET_4_6,
    messages=[{"role": "user", "content": "Explain Bedrock in one sentence."}],
    max_tokens=1024,
)

print(message.content[0].text)

bedrockrouter models list
bedrockrouter models list --provider anthropic
bedrockrouter models find "sonnet"

OpenAI Python

from openai import OpenAI

client = OpenAI(
    base_url="https://YOUR_PROXY_URL/v1",
    api_key="YOUR_API_KEY",
)

response = client.chat.completions.create(
    model="us.anthropic.claude-haiku-4-5-20251001-v1:0",
    messages=[{"role": "user", "content": "Explain Bedrock in one sentence."}],
)

print(response.choices[0].message.content)

Anthropic TypeScript

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

const client = new Anthropic({
  baseURL: "https://YOUR_PROXY_URL",
  apiKey: "YOUR_API_KEY",
});

const message = await client.messages.create({
  model: "us.anthropic.claude-haiku-4-5-20251001-v1:0",
  max_tokens: 1024,
  messages: [{ role: "user", content: "Hello, Bedrock!" }],
});

console.log(message.content);

Vercel AI SDK

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

const bedrock = createOpenAI({
  baseURL: "https://YOUR_PROXY_URL/v1",
  apiKey: process.env.BEDROCK_API_KEY,
});

const { text } = await generateText({
  model: bedrock("minimax.minimax-m2-v1:0"),
  prompt: "Write a haiku about distributed systems.",
});

boto3 (native Bedrock)

Use the AWS SDK directly with the bedrock-runtime client. Set endpoint_url to your proxy URL and pass your Bedrock API key as a bearer token.

import boto3, json

client = boto3.client(
    "bedrock-runtime",
    region_name="us-east-1",
    endpoint_url="https://YOUR_PROXY_URL",
    aws_access_key_id="x",       # placeholder — auth via bearer token
    aws_secret_access_key="x",   # placeholder
)

# Set the bearer token via the Bedrock API key env var
import os
os.environ["AWS_BEARER_TOKEN_BEDROCK"] = "YOUR_API_KEY"

response = client.converse(
    modelId="us.anthropic.claude-haiku-4-5-20251001-v1:0",
    messages=[
        {"role": "user", "content": [{"text": "Explain Bedrock in one sentence."}]}
    ],
)

print(response["output"]["message"]["content"][0]["text"])

BYOK Integrations

Bring your own vendor key (BYOK) to route traffic directly to OpenAI or Anthropic — with BedrockRouter handling telemetry and cost tracking. Use api_key for the third-party vendor secret and mach11_key for your BedrockRouter tracking key.

from bedrockrouter import BedrockOpenAI

client = BedrockOpenAI(
    api_key="YOUR_OPENAI_API_KEY",       # Your OpenAI secret key
    mach11_key="YOUR_BEDROCKROUTER_KEY"  # Your BedrockRouter tracking key
)

resp = client.chat.completions.create(
    model="gpt-4o-mini",
    provider="openai",   # Required — tells the proxy to forward to OpenAI
    messages=[{"role": "user", "content": "Say 'OpenAI routing is working!'"}]
)

print(resp.choices[0].message.content)

from bedrockrouter import BedrockRouter

client = BedrockRouter(
    api_key="YOUR_ANTHROPIC_API_KEY",    # Your Anthropic secret key
    mach11_key="YOUR_BEDROCKROUTER_KEY"  # Your BedrockRouter tracking key
)

msg = client.messages.create(
    model="claude-haiku-4-5",
    provider="anthropic",  # Required — tells the proxy to forward to Anthropic
    max_tokens=100,
    messages=[{"role": "user", "content": "Say 'Anthropic BYOK routing is working!'"}]
)

print(msg.content[0].text)

from bedrockrouter import BedrockRouter

client = BedrockRouter(
    api_key="YOUR_BEDROCKROUTER_KEY"  # Your BedrockRouter key — that's all
)

msg = client.messages.create(
    model="us.anthropic.claude-opus-4-7",  # Any Bedrock model ID
    max_tokens=512,
    messages=[{"role": "user", "content": "Hello!"}]
)

print(msg.content[0].text)

Supported models

BedrockRouter does not restrict which models you can call. Pass any model ID available in your AWS Bedrock region and the proxy forwards the request as-is. The models listed on the Models page are a curated catalog with pricing — they are not a whitelist.

# Anthropic Claude Haiku 4.5
model = "us.anthropic.claude-haiku-4-5-20251001-v1:0"

# Amazon Nova Lite
model = "us.amazon.nova-lite-v1:0"

# DeepSeek R1
model = "deepseek.r1-v1:5"

# MiniMax M2
model = "minimax.minimax-m2"

# Any other model from your Bedrock region
model = "<bedrock-model-id>"

Authentication

BedrockRouter uses virtual API keys issued from your Mach11 dashboard. Pass your key in the Authorization header on every request.

Authorization: Bearer sk_m11_xxxxxxxxxxxxxxxxxxxx

Keys are created and revoked from your dashboard. Set optional credit limits per key. The Lambda verifies each key using HMAC — no AWS credentials are exposed to clients.

Pricing & billing

BedrockRouter charges zero markup. You pay AWS Bedrock's standard on-demand rates directly via your AWS bill. The Lambda proxy is metered separately under AWS Lambda — typically a few cents per million requests.

FAQ