---
title: "Deploy Philter to the Cloud"
description: "Launch Philter (open source PII and PHI redaction) in your own AWS, Azure, or Google Cloud VPC. Pay-as-you-go through your existing cloud account."
url: "https://philterd.ai/philter/deploy/"
---

Cloud deployment

# Deploy Philter in your own cloud

Launch Philter’s open source PII and PHI redaction from your cloud’s marketplace. It runs on a virtual machine in your own account, billed with the rest of your cloud usage.

- **From $0.49/hour** on AWS Marketplace
- **~$350/month** running 24/7
- **Up to 18% off** with an annual contract

Choose your cloud

- **AWS Marketplace** EC2 instance `c7i.xlarge` recommended
   
   [Launch](https://aws.amazon.com/marketplace/pp/B07YVB8FFT?ref=_ptnr_mf_launch) [Quick start](https://philterd.ai/dev/start/aws/)
   
- **Azure Marketplace** Virtual machine `Standard_F4s_v2` recommended
   
   [Launch](https://azuremarketplace.microsoft.com/en-us/marketplace/apps/philterdllc1687189098111.philter?tab=Overview) [Quick start](https://philterd.ai/dev/start/azure/)
   
- **Google Cloud Marketplace** Compute Engine VM `c3-highcpu-4` recommended
   
   [Launch](https://console.cloud.google.com/marketplace/product/philterd-public/philter) [Quick start](https://philterd.ai/dev/start/google-cloud/)
   

Need a private offer or help with procurement? [Contact us](https://philterd.ai/philter/deploy/#).

## From launch to first redaction

1. **Subscribe and launch.** Pick a region, a size with at least 2 vCPU and 8 GB of RAM, and an SSH key. Private subnets are supported, including ones with no internet egress.
2. **Read your API key over SSH.** On first boot the virtual machine generates its own encryption key and first API key into `/opt/philter/.env`.
3. **Send text to `/api/filter`.** Philter serves HTTPS on port 8080.
4. **Tune policies for your data.** Choose how each type is handled: redact, mask, hash, encrypt, or replace. Start from the [policy library](https://philterd.ai/policies/) or the [User’s Guide](https://philterd.github.io/philter/).

Your first request

```
$ curl -sk https://<instance-ip>:8080/api/filter \
    -H "Authorization: Bearer $PHILTER_API_KEY" \
    -H "Content-Type: text/plain" \
    --data-binary "George Washington was a patient and his SSN is 123-45-6789."

{{{REDACTED-name}}} was a patient and his SSN is {{{REDACTED-ssn}}}.
```

The first request may take a few seconds while Philter loads.

## Frequently asked questions

Is there a free trial?

Yes. The free trial period may vary by cloud provider. During the free trial you are responsible for the underlying cloud hardware, but there is no charge for the Philter software.

How much will this actually cost me?

On AWS Marketplace, $0.49/hour for instance types with up to 4 vCPU, including the recommended `c7i.xlarge`; larger types are priced by size ($0.98/hour for 8 vCPU, $1.96/hour for 16 vCPU). Worked numbers for one 4 vCPU instance:

- **~$350/month** if running it 24/7
- **~$290/month** with the 365-day annual contract (~18% off)

The instance itself is billed separately by your cloud provider. See the instance type question below for sizing. Pricing on Azure and Google Cloud follows similar per-hour models; check each listing for the current published rate.

Can I run in an air-gapped VPC?

Philter needs no outbound internet access to redact. The default policy’s model ships in the image, and Philter connects out only to services you configure: a remote PhEye service, replacement generators, webhooks, or Phield. Once the virtual machine is running, you can move it to a subnet with no NAT gateway and no internet egress; your own network still needs to reach port 8080, and port 22 for SSH. Launching from the marketplace, and later updates, need internet access or an internal mirror. This is a common deployment pattern for healthcare and government workloads.

What instance type should I start with?

Philter needs at least 2 vCPU and 8 GB of RAM. We recommend 4 vCPU and 8 GB: `c7i.xlarge` on AWS, `Standard_F4s_v2` on Azure, or `c3-highcpu-4` on Google Cloud. Smaller types are for testing. Philter runs its models on CPU, so a GPU instance type adds cost without making it faster. One virtual machine runs one Philter; to run several behind a load balancer, see [High Availability](https://philterd.github.io/philter/latest/high_availability.html). Each cloud’s quick start covers sizing and launch: [AWS](https://philterd.ai/dev/start/aws/), [Azure](https://philterd.ai/dev/start/azure/), [Google Cloud](https://philterd.ai/dev/start/google-cloud/).

Can I bring my own license (BYOL)?

BYOL is not currently offered on the marketplace listings. If your team has procurement requirements that the marketplace doesn’t fit, [contact us](https://philterd.ai/philter/deploy/#). We can structure a direct engagement or private offer.

What support is included?

Email support from the Philterd team is included with marketplace deployments (typical response: within 48 hours). For named-contact SLAs, dedicated technical advisors, or hands-on engineering, see the [consulting services](https://philterd.ai/consulting/) or [contact the team directly](https://philterd.ai/philter/deploy/#).
