Overview

Infisical Relay is a secure routing layer that allows Infisical to connect to your private network resources, such as databases or internal APIs, without exposing them to the public internet. The relay acts as an intermediary, forwarding encrypted traffic between Infisical and your deployed gateways. This ensures that your sensitive data remains protected and never leaves your network unencrypted. With this architecture, you can achieve secure, firewall-friendly access across network boundaries, making it possible for Infisical to interact with resources even in highly restricted environments.

Before diving in, it's important to determine whether you actually need to deploy your own relay server or if you can use Infisical's managed infrastructure.

Do You Need to Deploy a Relay?

Not all users need to deploy their own relay servers. Infisical provides managed relay infrastructure in US/EU regions for Infisical Cloud users, which requires no setup or maintenance. You only need to deploy a relay if you:

• Are self-hosting Infisical
• Have a dedicated enterprise instance of Infisical (managed by Infisical)
• Require closer geographic proximity to target resources than managed relays provide for lower latency and reduced network congestion when accessing resources through the relay
• Need full control over relay infrastructure and traffic routing

If you are using Infisical Cloud and do not have specific requirements, you can use the managed relays provided by Infisical and skip the rest of this guide.

Deployment Steps

To successfully deploy an Infisical Relay for use, follow these steps in order.

To view all available flags and equivalent environment variables for relay deployment, see the [Relay CLI Command Reference](/cli/commands/relay).

**Inbound Connections Rules:**
| Protocol | Source             | Port | Purpose                          |
| -------- | ------------------ | ---- | -------------------------------- |
| TCP      | Gateways           | 2222 | SSH reverse tunnel establishment |
| TCP      | Infisical instance host (US/EU, other) | 8443 | Platform-to-relay communication  |

**Outbound Connections Rules:**
| Protocol | Destination                          | Port | Purpose                                    |
| -------- | ------------------------------------ | ---- | ------------------------------------------ |
| TCP      | Infisical instance host (US/EU, other) | 443  | API communication and certificate requests |

<AccordionGroup>
  <Accordion title="Token Auth">
    A one-time enrollment token (1h expiry) bootstraps the relay.

    <Tabs>
      <Tab title="Linux (Production)">
        ```bash
        sudo infisical relay systemd install \
          --name=<relay-name> \
          --enroll-method=token \
          --token=<enrollment-token> \
          --domain=<your-infisical-domain>
        sudo systemctl start infisical-relay
        ```
      </Tab>
      <Tab title="Foreground">
        ```bash
        infisical relay start \
          --name=<relay-name> \
          --enroll-method=token \
          --token=<enrollment-token> \
          --domain=<your-infisical-domain>
        ```
      </Tab>
    </Tabs>
  </Accordion>
  <Accordion title="AWS Auth">
    The host must have AWS credentials whose principal matches your allowlist. The relay re-authenticates via STS on every start.

    <Tabs>
      <Tab title="Linux (Production)">
        ```bash
        sudo infisical relay systemd install \
          --name=<relay-name> \
          --enroll-method=aws \
          --relay-id=<relay-id> \
          --domain=<your-infisical-domain>
        sudo systemctl start infisical-relay
        ```
      </Tab>
      <Tab title="Foreground">
        ```bash
        infisical relay start \
          --name=<relay-name> \
          --enroll-method=aws \
          --relay-id=<relay-id> \
          --domain=<your-infisical-domain>
        ```
      </Tab>
    </Tabs>
  </Accordion>
</AccordionGroup>

<Warning>
  The systemd install command requires Linux with root/sudo privileges.
</Warning>

<Info>
  Token-method enrollment tokens are single-use and expire after 1 hour. If the token expires before deployment, click **Show deploy command** again on the relay detail page to generate a new one.
</Info>

<Info>
  You can safely re-run the same command to restart the relay. The CLI detects the token has already been used locally and skips enrollment automatically.
</Info>

Frequently Asked Questions

• Client-to-Gateway mTLS (via TLS-pinned tunnel): Clients connect via a proxy that establishes a TLS-pinned tunnel to the gateway; mTLS between the client and gateway is negotiated inside this tunnel, encrypting all application traffic
• SSH tunnel encryption: The mTLS-encrypted traffic is then transmitted through SSH reverse tunnels to relay servers
• Double encryption: Traffic is encrypted twice - once by client mTLS and again by SSH tunnels
• Relay only routes traffic: The relay server only routes the doubly-encrypted traffic without access to either encryption layer

The relay infrastructure is designed as a secure routing mechanism where only the client and gateway can decrypt the actual application traffic.

• Dedicated resources: Full control over relay infrastructure and performance
• Lower latency: Deploy closer to your gateways for optimal performance
• Compliance: Meet specific data routing and compliance requirements
• Custom network policies: Implement organization-specific network configurations
• Geographic proximity: Reduce network congestion and improve response times to access resources
• High availability: Deploy multiple relays for redundancy and load distribution

Organization-deployed relays give you complete control over your secure communication infrastructure.

Platform cannot connect to relay:

• Check firewall rules allow inbound TCP with TLS on port 8443
• Test connectivity: openssl s_client -connect <relay-ip>:8443

Test network connectivity:

# Test outbound API access from relay. Replace URL with your Infisical instance if self-hosted
curl -I https://app.infisical.com

# Test TCP with TLS port from platform
openssl s_client -connect <relay-ip>:8443

• Gateway reconnection: Gateways will automatically attempt to reconnect when the relay comes back online
• Service interruption: While the relay is down, the Infisical platform cannot reach gateways through that relay. As a result, any secrets or resources accessed via those gateways will be temporarily unavailable until connectivity is restored.
• Multiple relays: Deploy multiple relay servers for redundancy and high availability
• Automatic restart: Use systemd or container orchestration to automatically restart failed relay services

For production environments, consider deploying multiple relay servers to avoid single points of failure.