Overview
RouterBench is an enterprise LLM gateway that gives your teams a single, OpenAI-compatible endpoint for accessing 100+ AI model providers - including Amazon Bedrock, OpenAI, Anthropic, and Google Vertex AI -with automatic smart routing to the most cost-effective, capable model for every request. This engagement deploys RouterBench directly inside your own AWS account (using services such as Amazon EC2, Amazon ECS/Fargate, Amazon RDS, and Amazon ElastiCache), so your prompts, responses, and API keys never leave your network. Our team handles architecture design, deployment, configuration of virtual API keys, budgets and rate limits, SSO/RBAC integration, guardrails and PII redaction, and full observability - followed by knowledge transfer so your team can operate it independently. RouterBench includes per-team virtual keys, spend budgets and token quotas, automatic provider fallback, prompt/response caching, audit logging, and a management dashboard -replacing scattered point integrations with a single governed layer for all LLM traffic.
Highlights
- Deployed entirely inside your own AWS account - your prompts, responses, and provider API keys never leave your network.
- Smart routing across 100+ LLM providers (including Amazon Bedrock, OpenAI, and Anthropic) automatically picks the most cost-effective, capable model per request.
- Enterprise governance out of the box: virtual API keys, per-team budgets and token quotas, SSO/RBAC, audit logs, and guardrails.
Details
Introducing multi-product solutions
You can now purchase comprehensive solutions tailored to use cases and industries.
Features and programs
Financing for AWS Marketplace purchases
Pricing
Dimension | Description | Cost/12 months |
|---|---|---|
RouterBench License | License to deploy and use RouterBench in your own AWS account. | $0.001 |
Vendor refund policy
Subscription charges are non-refundable once a billing period has begun, except in the following cases: you were charged in error, or a defect prevents you from using the software and we cannot resolve it within 10 business days of a support request.
To request a refund, email support@routerbench.com within 30 days of the charge with your AWS account ID and order ID. Approved refunds are processed through AWS Marketplace and issued by AWS to your original payment method.
How can we make this page better?
Legal
Vendor terms and conditions
Content disclaimer
Delivery details
RouterBench single-instance deployment
This template deploys RouterBench as a single-instance stack in your own AWS account and VPC. All data and all traffic stay inside your account.
WHAT IT CREATES
- An EC2 instance (c7i or m7i) running the full RouterBench stack in Docker: API gateway, control-plane API, customer console, admin app, background worker, PostgreSQL, Redis and a Caddy reverse proxy with automatic TLS.
- An encrypted EBS data volume, separate from the root disk, holding the database and all persisted state. The volume survives instance replacement, and an existing volume can be re-attached at launch to restore an instance.
- A Data Lifecycle Manager policy that takes daily snapshots of the data volume, with a retention period you set.
- An Elastic IP, so the address survives a stop/start.
- A security group allowing HTTPS (TCP and UDP 443) and HTTP (TCP 80, for certificate issuance) from a CIDR you choose. SSH is closed unless you supply both SshCidr and KeyName.
- An IAM role and instance profile granting only Session Manager access, CloudWatch logging and the snapshot permissions the stack needs.
- Three Route 53 A records (api, app, admin) if you supply a hosted zone ID. Otherwise the stack outputs the records for you to create yourself.
CloudFormation Template (CFT)
AWS CloudFormation templates are JSON or YAML-formatted text files that simplify provisioning and management on AWS. The templates describe the service or application architecture you want to deploy, and AWS CloudFormation uses those templates to provision and configure the required services (such as Amazon EC2 instances or Amazon RDS DB instances). The deployed application and associated resources are called a "stack."
Version release notes
Model catalogue updated: 24 provider mappings and 14 models retired upstream were removed, verified against each provider's API. Automatic model selection no longer considers embedding, speech, image or video models, which could previously be chosen for a chat request and fail at the provider. A provider key rejected only for billing or quota is now saved with a warning instead of being refused.
Additional details
Usage instructions
BEFORE YOU LAUNCH
- A domain you control. Three hostnames are served under a base domain you choose, e.g. example.com: api.example.com - the API (/v1/...) app.example.com - the console admin.example.com - the admin app
- A VPC and a public subnet, so Let's Encrypt can reach the instance on port 80 to issue certificates. If the instance cannot be public, set TlsMode = internal (browsers will show a certificate warning).
- An administrator email address. This becomes the first account.
LAUNCH
-
Launch the CloudFormation template from this listing.
-
Fill in Base domain and Admin email. If the domain is hosted in Route 53, supply its hosted zone ID and the DNS records are created for you.
-
Wait for the stack to reach CREATE_COMPLETE (about 5-10 minutes).
-
If you did not supply a hosted zone, create the DNS records shown in the DnsInstructions stack output: three A records, or one wildcard *.example.com, pointing at the PublicIp output. Certificates are issued automatically within a few minutes of those records resolving.
-
Open the ConsoleUrl stack output and sign in with your admin email. A strong password is generated on the instance at first boot; it is not stored in the AMI and differs on every instance. Read it once over Session Manager (EC2 console, Connect, Session Manager):
sudo cat /var/lib/routerbench/data/config/initial-admin-password
RouterBench then requires you to set a new password before any other page or API call will work. Delete that file once you have signed in.
All other credentials (database password, cache password, session signing key, encryption key) are generated randomly on the instance at first boot and are never baked into the AMI.
FIRST STEPS
-
Add provider API keys in the admin app, then enable models in the catalogue.
-
Create an API key in the console and call the gateway:
curl https://api .<your-domain>/v1/chat/completions
-H "Authorization: Bearer <your-key>"
-H "Content-Type: application/json"
-d '{"model":"auto","messages":[{"role":"user","content":"hello"}]}' -
Invite your team from the console, or configure single sign-on.
ACCESSING THE INSTANCE
Use Session Manager: EC2 console -> Connect -> Session Manager. The instance role permits it. SSH is closed unless you set SshCidr and KeyName at launch, and password login is disabled.
sudo systemctl status routerbench cd /opt/routerbench && sudo docker compose ps sudo docker compose logs -f api gateway sudo cat /var/log/routerbench-firstboot.log
CONFIGURATION
Settings live in /opt/routerbench/.env on the data volume. Edit it as root, then run: sudo systemctl restart routerbench
Email (invites, password resets): set SMTP_HOST, SMTP_PORT, SMTP_USER, SMTP_PASS and SMTP_FROM_EMAIL. Until these are set, no email is sent. Single sign-on (OIDC - Entra ID, Okta, Google Workspace): set OIDC_DISCOVERY_URL, OIDC_CLIENT_ID, OIDC_CLIENT_SECRET and SSO_ENABLED=true. Register this redirect URI with your provider: https://api .<your-domain>/auth/oauth2/callback/sso
Do not add PORT to the .env file.
BACKUP AND RESTORE
All state lives on the encrypted EBS data volume, which the stack snapshots daily. To restore, launch a new stack and pass the existing volume ID as ExistingDataVolumeId.
Support
Vendor support
Support is provided via email at support@routerbench.com , with response times of 1 business day for standard requests and same-day response for production-impacting issues under an active engagement. Each engagement includes architecture documentation, operational runbooks, and a knowledge-transfer session so your team can operate the deployment independently. Extended/premium support plans are available on request as part of your private offer.
AWS infrastructure support
AWS Support is a one-on-one, fast-response support channel that is staffed 24x7x365 with experienced and technical support engineers. The service helps customers of all sizes and technical abilities to successfully utilize the products and features provided by Amazon Web Services.