What it costs to run
SecureMind itself is free. What you pay for is where it runs and the models you connect. On your own computer that is nothing. In your own AWS account, a small install with no traffic comes to well under a dollar a month.
On your own computer
$0. Docker runs the app, the database, sign-in and the local models on your machine, so it uses your own electricity and disk space and nothing else. The only cost appears if you choose to connect a paid model provider, and that is optional.
In your AWS account
Estimated monthly cost for the install as it is built today, in us-east-1. Prices were checked in October 2026. The last column is what each service costs with no one using it.
Parameter Store
$0.05
Holds the encryption key and setup code, and the vendor keys an admin saves. Encrypted at rest.
Standard parameters are free. The vendor-keys parameter uses the Advanced tier, at $0.05 per parameter per month.
Secrets Manager would charge $0.40 for each of these, so the stored values use Parameter Store instead.
CloudWatch Logs
$0.00
Keeps Lambda logs for 14 days.
$0.50 per GB ingested. The first 5 GB a month are free.
A small install writes far less than 5 GB.
Amplify Hosting
$0.00 to $0.10
Builds and serves the web app.
Builds $0.01 a minute. Storage $0.023 per GB-month. Data served $0.15 per GB. Server rendering $0.30 per million requests plus $0.20 per GB-hour.
Auto-build is off, so builds run only when you start one, at a few minutes each. New AWS accounts also get a monthly free allowance for 12 months.
API Gateway
$0.00
Front door for the chat streaming endpoint.
$3.50 per million requests.
Nothing is charged while nobody chats.
Lambda
$0.00
Runs the chat function and a small warm-up function (Arm, Node.js 22).
$0.20 per million requests plus $0.0000133334 per GB-second. The free allowance is 1 million requests and 400,000 GB-seconds every month.
The chat function has 1.5 GB of memory and runs only while a reply streams.
DynamoDB
$0.00
Stores memory fragments and the vector index used for Knowledge search.
On demand: $0.625 per million writes and $0.125 per million reads. Storage $0.25 per GB-month, with the first 25 GB free. Vector search is billed by data processed.
No capacity is reserved, so an idle table costs nothing beyond storage.
Cognito
$0.00
Invite-only sign-in.
The first 10,000 monthly active users are free on the default tier.
S3
$0.00
Profile pictures, and Terraform's state file.
About $0.023 per GB-month.
A few kilobytes to a few megabytes, so it rounds to cents at most.
IAM, ACM certificates
$0.00
Roles, permissions and HTTPS certificates.
No charge.
Total, idle
About $0.05
The install itself, with nobody using it. Plus a few cents each time you start a build.
Route 53 (optional)
+ $0.50*
DNS for your own domain name.
$0.50 per hosted zone per month, plus $0.40 per million queries.
Only if you use your own domain. Without one, the Amplify address works and this is $0.
* The install does not create or need a domain name. If you point your own domain at it, AWS DNS (Route 53) adds about $0.50 a month for the hosted zone, and registering the name is a separate yearly fee of roughly $10 to $20.
What it costs as it gets used
Idle
About $0.05 to $0.15
a month, before model fees
Installed and not used, or used a few times a month.
The vendor-keys parameter, plus a few cents for each build you start. A domain of your own adds about $0.50.
A few people
About $0.10 to $0.70
a month, before model fees
5 people, about 2,000 chat messages a month.
The extra traffic stays inside the free allowances for Lambda, Cognito and logs. API Gateway adds under a cent. The rest is the idle cost.
Busy
About $10 to $15
a month, before model fees
50 people, about 50,000 chat messages a month.
Lambda becomes the main cost, because the function stays running while each reply streams. Assumes about 15 seconds per reply.
These are estimates
A real bill
This is the author's own install, in a dedicated AWS account, with a custom domain and light use. The figures are from AWS Cost Explorer, before tax.
| Service | August | September | October, to the 18th |
|---|---|---|---|
| Route 53 | $0.51 | $0.52 | $0.51 |
| Amplify | $0.00 | $0.00 | $0.20 |
| Secrets Manager | $0.00 | $0.00 | $0.34 |
| S3 | $0.00 | $0.00 | under $0.01 |
| Lambda, API Gateway, DynamoDB, Cognito | $0.00 | $0.00 | $0.00 |
August and September were the domain only, because the app was not installed yet. October is the first month with the app running, and it includes a lot of testing. The $0.20 for Amplify is about 20 build minutes. Lambda, API Gateway, DynamoDB and Cognito stayed inside their free allowances. The secrets line is the one this page used to show at $1.20 a month, and it is gone in the current code. A few cents of other charges in the same account come from unrelated resources.
Not included
- Model usage. Amazon Bedrock and every vendor you connect (Anthropic, OpenAI, Gemini, Groq, Grok) charge per token. This is usually the largest cost once people start chatting, and it varies a lot by model. The app shows an estimate on each reply and an admin can set a monthly cap.
- Your domain name. Registering one costs roughly $10 to $20 a year at a registrar. It is optional.
- Data transfer and taxes. Traffic to your browser is small for a text app. Taxes depend on where you are.
- Free-tier credits. A new AWS account gets credits and a free tier for a while, which lowers the first months. The table above assumes you pay full price.
Why it is built this way
This is a personal tool meant to sit idle most of the time, so the design question was how to make an idle install cost almost nothing.
Pay for use, not for being switched on
Lambda, API Gateway, DynamoDB on demand and Cognito charge only when something happens. There is no server, database instance or reserved capacity running around the clock.
No network appliances
There is no custom VPC, no NAT gateway and no load balancer. Those are the usual reasons a small AWS project costs $30 to $100 a month before anyone uses it.
A database that scales to zero
DynamoDB on demand holds the data and also does the vector search, so there is no separate search cluster or vector database to pay for.
Small, cheap Lambda functions
Both functions run on Arm, which is cheaper per second. The warm-up function is tiny, and logs are kept for only 14 days.
Builds happen when you say so
Auto-build is off. Amplify builds only when you start one (or when you turn on the optional GitHub deploy), so a stray push does not spend build minutes.
No model bill unless you want one
The app does not pay for a model on your behalf. You connect your own keys or Bedrock, so model cost is yours to control. An admin can set a monthly cap in the app.
Everything can be removed
One command deletes what was created, so a trial does not turn into a standing bill.
Keeping an eye on it
- Create an AWS Budget with an alert, for example at $5 a month, so a surprise reaches your email first.
- Look at Cost Explorer, grouped by service, after the first month and compare it with this page.
- Use the cost cap in Settings to limit model spending.
- When you are done, remove everything with
npm run infra:down -- --yes. The steps are on the Install page.
Prices change. The AWS pricing pages are the final word. If a figure here looks out of date, open an issue.