Site-to-site installation guide for end customers

Overview

Why am I being asked to install ngrok?

Your vendor wants to create a secure persistent connection between your network and theirs, which allows them to access and take action on your services and data. We call this site-to-site connectivity.

Your vendor will create the required configuration files and deployment strategy and share them with you directly. You must contact with your vendor to request changes to that configuration, including those based on the content of this document.

What is ngrok?

ngrok is a universal gateway—an all-in-one reverse proxy, API gateway, Kubernetes ingress, firewall, and global load balancer to deliver apps and APIs.

Why are we using ngrok instead of other solutions, like a VPN or opening ports?

ngrok is simpler and more secure than these other solutions for a few reasons:

• ngrok does not require you to open inbound ports in your network to the public internet.
• ngrok’s agent is infrastructure-agnostic, operating the same way in any cloud, region, or environment.
• The agent is simpler to configure, deploy, and maintain than VPNs or VPC peering by collapsing many networking components, like load balancing, encryption, certificate management, and authentication into a unified platform.
• ngrok’s network includes DDoS protection and global acceleration for all connections through the Global Server Load Balancer (GSLB).

ngrok's solution supports multiple models for TLS encryption, including end-to-end encryption. Learn more about how your vendor can configure ngrok's encryption: How does ngrok’s traffic encryption work?

Who else uses ngrok to provide access to private services?

Organizations worldwide trust ngrok for site-to-site connectivity, unified ingress, device gateways, and developer productivity. Our customers include Twilio, Okta, Zoom, Microsoft, Zendesk, Cyera and many more.

Databricks, the leading unified lakehouse platform for data, analytics, and AI, uses ngrok for its site-to-site connectivity across all of its customers.

You can learn more about our customers and read case studies about their successes on our customers page.

Who is ngrok? Where is it based?

ngrok was first released in 2013 as an open-source project by Alan Shreve, who founded ngrok as a company in 2015 and remains our CEO. Our executive team has extensive experience at companies building distributed systems and winning organization cultures.

In December 2022, ngrok closed $50 million in Series A funding with Lightspeed Venture Partners and Coatue.

ngrok has a primarily US-based workforce. More information on the geographic distribution of full time employees and contractors can be found at our Trust Center.

Can I trust ngrok as a company?

More than 7 million developers use ngrok. We’re recommended by category leaders including Twilio, GitHub, Okta, Microsoft, Zoom, and Shopify. We operate a global network and we have handled over 100 trillion total requests.

ngrok is SOC 2 Type 2 compliant, which certifies that our security processes and operations meet AICPA's criteria for security. We are also CCPA, EU-US DPF, and GDPR compliant.

How ngrok works

How does ngrok work?

ngrok has three primitives: the network, the agent, and the dashboard.

• The network is a globally distributed infrastructure that accepts traffic to endpoints, applies policies, and routes connections to the appropriate agent.
• The agent is a lightweight program with zero runtime dependencies, which forwards requests from your vendor to your upstream services and data.
• The dashboard is a SaaS web app for configuring accounts, endpoints, and access policy. Depending on the architecture and arrangement with your vendor, you may or may not have access to an ngrok dashboard.

In a generic site-to-site connectivity architecture, these pieces come together to do the following:

1. The ngrok agent in your network connects to one or more of your upstream services.
2. The ngrok agent connects out to the ngrok network at the default agent ingress address at connect.ngrok-agent.com (or a custom one).
3. The ngrok network creates an endpoint like your-company.your-vendor.com, which then points to your upstream service.
4. Your vendor sends requests from their services to the endpoint, which are then routed through to the agent and finally your upstream service.

What is ngrok’s network architecture?

We run ngrok’s primary network and services on AWS.

This architecture includes multiple Points of Presence (PoPs) worldwide that run our services, route customer traffic, and make up our always-on Global Server Load Balancer (GSLB).

The GSLB, which is active by default for all ngrok endpoints and tunnels, enhances the resiliency of the connection between your services and your vendor’s with DDoS protection and geo-failover that automatically routes traffic to the lowest-latency PoP in the event of downtime or service interruption. If you have specific data residency requirements, like those located in the EU, you and your vendor can configure ngrok to use a specific region in Europe or pin all traffic to a specific PoP.

There are many possible architectures for setting up a site-to-site network based on the shared requirements of you and your vendor. Your vendor will work with you to find a mutually secure and reliable architecture.

How does ngrok handle data residency requests?

Your vendor can configure your site-to-site architecture to match your data residency needs.

They will first configure the agent to use a PoP in one of our supported regions, then work with you to set up appropriate DNS to route all connections through the same data plane.

Our regional data planes are located in Australia (Sydney), Europe (Frankfurt), India (Mumbai), Japan (Tokyo), South America (São Paulo), United States (California and Ohio).

Is ngrok a dedicated instance or multi-tenant?

ngrok is a multi-tenant application with services shared across our customer base, including your vendor and your site-to-site architecture.

Security

How can I ensure ngrok does not pose a security risk to my infrastructure?

We recommend you begin by exploring our Security, Privacy, and Compliance page, followed by our Trust Center.

Your vendor can implement multiple security practices, including:

• Prevent unauthorized usage of ngrok outside of the site-to-site connectivity with your vendor.
• Create a single account tenant, enforce Account Domain Controls to route all ngrok users with an email address @your-company.com into it, and enable role-based access controls to limit those accounts.
• Create bot users to create authtokens not tied to any individual user, and use a different, scoped authtoken for each agent.
• Apply Access Control Lists (ACLs) to authtokens to control where ngrok agents can be created within your network.
• Enable the IP restrictions action on your endpoint(s) to allow only trusted connections from your vendor.
• Enforce basic auth, OAuth, OpenID, SAML, JWT, or mTLS authentication on your endpoint(s).

The configuration and operation of these security practices will be handled by your vendor.

How does ngrok’s traffic encryption work?

The agent always connects to the ngrok network via TLS. We support three encryption models based on where TLS is terminated:

1. TLS termination at the ngrok network. This is the default behavior when using HTTPS tunnels, where ngrok manages TLS certificate generation and renewal for both you and your vendor.
2. TLS termination at the ngrok agent. Often referred to as zero-knowledge TLS, this model encrypts your TLS tunnels end-to-end, using filesystem paths to your TLS certificate and key, without requiring you to reconfigure your upstream service for TLS termination.
3. TLS termination at your upstream service. This model implements end-to-end zero-knowledge encryption, where the ngrok network forwards unterminated connections through to your upstream application.

Is my data end-to-end encrypted?

Yes—if you vendor configures ngrok to terminate TLS at the agent or your upstream service using the second or third models listed above.

Contact your vendor if you’re unsure how TLS termination is managed in your site-to-site connectivity architecture.

Can ngrok see my traffic?

No—if your vendor configures ngrok to terminate TLS at the agent or your upstream service.

In all encryption models, the ngrok agent cannot see the traffic it forwards on to your upstream service.

If your vendor requests HTTP tunnels for your site-to-site connectivity use case and also uses the Traffic Inspector feature, then ngrok can see and will store up to 90 days of metadata about each request and response. If your vendor enables Full Capture mode, ngrok also stores the full request and response parameters, headers, and bodies for each traffic event.

Can I set up deep packet inspection or observe what data transmits through ngrok?

Yes. Your vendor can configure your site-to-site connectivity architecture in a few ways to enable this functionality. They can:

1. Configure your ngrok agent to trust a specific root certificate you own on the host’s OS or a specific PEM file on disk instead of the trusted certificate root for the ngrok network. You can then use a proxy for deep packet inspection.
2. Require the outbound connection to connect.ngrok-agent.com:443 go through a proxy capable of inspection.
3. Set up a bypass rule for connect.ngrok-agent.com:443 (or a custom agent ingress address) to not perform TLS inspection.
4. Help you set up software between the ngrok agent and your upstream service, or the ngrok agent and the ngrok network, to see what’s transmitted through ngrok.

In any case, your vendor is responsible for configuring these inspection strategies for your site-to-site architecture.

How can I prevent ngrok from being used for any purpose other than in site-to-site connectivity with my vendor?

The best way to disallow other uses of ngrok on your network is working with your vendor to specify a custom agent ingress address.

Your vendor will configure their DNS to issue certificates for the custom address, such as your-company.us.connect.your-vendor.com:443. Then they’ll work with you to reconfigure your ngrok agents to utilize the custom ingress address.

Your vendor can also work with ngrok to provision dedicated IPs for their custom ingress address.

At this point, you can block the default agent ingress address at connect.ngrok-agent.com:443, which all agents use by default to connect outbound to ngrok’s network. This address resolves to a dynamic set of IP addresses, and blocking it at your network prevents any usage outside of this site-to-site connectivity use case in partnership with your vendor, such as developers utilizing ngrok to tunnel local development environments to the public internet.

You should also block the ngrok-managed public domains:

• ngrok.app
• ngrok.dev
• ngrok-free.app
• ngrok-free.dev
• ngrok.io
• ngrok.pizza
• ngrok.pro

Work with your vendor to design an architecture that prevents unauthorized use and uses the appropriate encryption model between your services.

How does ngrok mitigate abuse of its platform?

ngrok has a multi-pronged strategy for combating malicious use of our network, including automated systems that flag suspicious activity in combination with human moderation.

We work with authorized third-party security vendors, who report suspected abuse to abuse@ngrok.com or our abuse API. Our abuse response team investigates these reports, collaborates with these third parties, and proactively shuts down attacks.

We employ various passive abuse prevention techniques, such as restricting product usage for free and unverified ngrok accounts. We also use our IP Intelligence system to prevent signups from IPs on trusted blocklists, from specific regions, and those using anonymous proxies. Finally, our global firewalls protect against unauthorized access and DDoS attacks.

See our abuse page for more details.

Privacy and compliance

What compliance certifications does ngrok maintain?

ngrok complies with SOC 2, GDPR, EU-US DPF, and CCPA.

You can request our SOC 2 report and view other documents about ngrok’s security measures, like annual penetration testing, on our Trust Center.

What data does ngrok have access to, and how long is it stored?

ngrok’s access to your data depends on the encryption model specified by your site-to-site connectivity architecture—reach out to your vendor for more details.

In all cases, ngrok stores information about the machine where you run your agent, such as its IP address, operating system, CPU architecture, and anonymized details about the environment.

ngrok also logs changes to account configurations, such as agent start/stop, API key creation, and more. You can see the list in our Event Store documentation.

We retain data about your site-to-site connection at the direction of your vendor, who can also delete data at any time.

Read our primer on data at ngrok for more details.

Where does data transmitted through ngrok go?

ngrok is split into a control plane and data plane. The control plane is responsible for handling API requests and is the source of truth for account and user data. The data planes are responsible for hosting the tunnels and routing the customer traffic. ngrok also leverages several third party providers when handling customer data.

Control Plane: Any information stored by ngrok would reside in the control plane. This is hosted out of a US-based AWS region.

Data Plane: Customer traffic runs through ngrok’s regional data planes, located in Australia (Sydney), Europe (Frankfurt), India (Mumbai), Japan (Tokyo), South America (São Paulo), United States (California and Ohio)*.

By default, ngrok will route traffic via the most low latency route. Customers can control which regional data plane their tunnels connect to via the agent configuration and which regional data plane their customers connect into via DNS. If a region is specified, customer traffic will only be routed through that region.

What security practices does ngrok follow?

ngrok uses the shared security responsibility model where we are responsible for the security of our network, and we deliver features your vendor can use to configure and secure your site-to-site connectivity architecture.

Our fundamental security practices include access control via an identity provider, change management, full encryption at rest, and much more.

See our Security, Privacy, and Compliance and Trust Center pages for details.

Operations and agent deployment

Where can I run the ngrok agent and what are the ngrok agent’s system requirements?

The ngrok agent runs on Linux, Windows, and macOS systems and most CPU architectures. See our supported platforms documentation for details about supported CPU architectures and resource requirements.

We also distribute the agent as SDKs, a Docker container, and a Kubernetes Operator.

Your vendor will work with you to find the right form factor for your site-to-site connectivity architecture.

How do I manage the lifecycle and maintenance of the agent?

Your vendor is responsible for helping you configure and maintain your agent(s).

They might configure your ngrok agent to run in the background as a native operating system service. This functionality works on all Linux, Windows, and macOS systems, and once installed, the ngrok service starts at boot-time, automatically restarts after crashes, and sends logs to your system’s native logging service.

If they suggest deploying with Docker, they will likely recommend Docker’s restart policies or systemd integration; with our Kubernetes Operator, they will recommend the container restart policy is set to Always.

ngrok releases security and feature updates through all our installation channels and package managers. These updates are not automatic. Do not update your ngrok agents manually—your vendor will work with you on when and how to update your ngrok agent(s) because the process depends on how they asked you to install them.

• Package manager (brew or apt): Get updated agent versions through the same channel (e.g. brew update && brew upgrade package_name or apt update && apt install ngrok).
• Binary from our downloads page: Follow the same process again or update directly from your CLI with ngrok update.
• Agent SDK: Use the package manager for your language or framework (e.g. go get -u)
• Kubernetes Operator: Use helm upgrade.

Can I run the ngrok agent in a DMZ?

Yes. There are two requirements in this case:

1. The system(s) in your DMZ must support the general system requirements.
2. The upstream service your vendor wants to connect to with ngrok must also be running in your DMZ.

How does the agent connect to the ngrok network?

By default, the ngrok agent attempts to connect to the default agent ingress address at connect.ngrok-agent.com:443, which resolves to a dynamic list of IP addresses. You can view the full list of addresses and IPs available for the agent to connect to in our AWS-hosted tunnel.json file.

This connection includes the following steps:

1. DNS resolution to discover the appropriate connection address.
2. Verification of the TLS certificate returned by the ngrok network, signed by ngrok’s root certificate authority, against the certificate authorities bundled into the agent.
3. A Certificate Revocation List (CRL) check to crl.ngrok-agent.com/ngrok.crl:80 to ensure you never use an ngrok network certificate that has been revoked.

After the agent connects to the network, it also sends heartbeat messages to validate the liveness of the connection, and checks for updates on startup.

How does the agent continue running if my server loses connection or times out?

The ngrok agent utilizes a heartbeat that attempts to reach ngrok’s network every 10 seconds. If the agent doesn’t receive a response within the tolerance window (15 seconds by default), it terminates the existing connection and attempts to reconnect.

This mechanism allows your site-to-site connectivity to automatically reestablish after packet loss, dynamic IP changes, or complete network outages.

Your vendor can configure both the heartbeat interval and tolerance per agent.

How does using ngrok, and the availability of your platform, affect our own?

ngrok is only utilized for connectivity to the agent. It won’t affect the rest of your platform in any way. The ngrok platform is designed for high availability. It operates out of 8 separate isolated regions and multiple availability zones in each region within the AWS network. Your traffic will take the lowest latency path to the ngrok servers and will fail over to a secondary region in the case of network outages or infrastructure failures.

Who should we contact for support?

For issues or concerns around configuring and securing your site-to-site architecture, please contact your vendor.