Compare
hotcfg vs S3 + CloudFront and AWS AppConfig
hotcfg is versioned JSON config at the edge, with no buckets, invalidations or IAM policies to run. How it compares with S3 + CloudFront and AWS AppConfig.
If you're here because…
- Nothing stops a bad upload. S3 stores whatever bytes it's given, so a typo, or a string where your app expects an integer, goes live the moment it's uploaded.
- Half your users still get the old file, because CloudFront kept the cached copy and the invalidation was slow or forgotten.
- Rolling back means digging through object versions and running another invalidation while production is broken.
- AppConfig is a lot of setup for one JSON file: an application, an environment, a configuration profile, a deployment strategy and IAM permissions, then a two-step session API instead of a plain GET.
- Your readers are apps, not AWS servers. Phones, browsers and game clients shouldn't need the AWS SDK, credentials or signed requests to read their config.
…hotcfg replaces that glue with one URL and a schema.
Why hotcfg
A schema refuses broken uploads before they go live
Attach a JSON Schema to a bin, and every publish, promote and rollback is checked against it before it's stored. A wrong type or a missing key is refused with the file, the line and the rule it broke:
$ hotcfg push flags.json --bin mobile-flags --env staging
error: flags.json does not match the schema of bin mobile-flags (1 error)
flags.json:3 /limits/max_upload_mb
rule: /properties/limits/properties/max_upload_mb/type
message: must be an integer, not a stringAppConfig can also check a JSON Schema, or run a Lambda function, when a deployment starts. hotcfg runs the check on every push, promote and rollback, and in the dashboard's editor as you type.
The edge cache is purged for you
Serving JSON from S3 through CloudFront means a bucket policy, CORS headers, Cache-Control metadata, and an invalidation after every upload. Public bins in hotcfg are served from Cloudflare's edge with none of that set up: every publish, promote and rollback purges the cached copy, and if hotcfg's servers are ever unreachable, the edge keeps serving the last copy for up to 7 days.
A plain HTTPS GET
Reading a public bin is one unauthenticated request, with ETag and 304 Not Modified. No AWS SDK, no signed requests, no session token to carry between calls:
const config = await (await fetch("https://api.hotcfg.run/v1/acme/mobile-flags")).json();AppConfig is built for EC2, ECS and Lambda: clients start a configuration session, then poll for the latest configuration with a token that changes on every call, or run the AppConfig Agent beside the app. A mobile app or a browser would need AWS credentials to do that. The recommended fetching code for hotcfg adds a 2-second timeout and a fallback to the last good copy, in one short file for each common language.
Checked in CI, promoted with a diff
Keep the config file in your repository, and the hotcfg GitHub Action checks every pull request against the bin's schema. On merge, push to staging; when it looks right, promote that exact version to production after reading the diff:
- uses: hotcfg/action@v1
with:
command: validate
file: config/flags.json
bin: acme/mobile-flags
api-key: ${{ secrets.HOTCFG_API_KEY }}On Pro, give CI a key that can only write one bin's staging environment, with no IAM policy or OIDC trust to write.
Types from the schema
S3 and AppConfig hand your app bytes; keeping its structs in step with the file is up to you. hotcfg types turns the bin's schema into the types your app decodes config into, in TypeScript, Swift, Kotlin, Java, Python, PHP, Ruby, Go, GDScript or C#:
hotcfg types --lang swift --bin mobile-flags -o Config.swiftIn CI, --check fails the build when the committed file is out of date.
Notifications and an audit log, built in
With S3, telling your team that config.json changed, and who changed it, means wiring S3 event notifications to EventBridge, Lambda or SNS. On Pro, a bin posts every publish, promote, rollback and unpublish to Slack, Discord, Microsoft Teams, Google Chat, Mattermost or Telegram, with who did it and a link to the diff, or sends a signed JSON event to your own endpoint. See Webhooks.
Why pay when S3 costs pennies?
A few kilobytes of JSON in S3 cost almost nothing to store, and AppConfig charges $0.0000002 per configuration request plus $0.0008 per configuration received. The cost of rolling your own is engineering time: validation scripts, cache headers and invalidations, notification plumbing and a rollback runbook, all kept working. hotcfg is free for 3 public bins, and Pro is a flat $19 a month per organization, with unlimited reads and unlimited seats.
At a glance
| hotcfg | S3 + CloudFront | AWS AppConfig | |
|---|---|---|---|
| Validation | A JSON Schema checked on every publish, promote and rollback, and in CI | None built in | JSON Schema or Lambda validators, checked when a deployment starts |
| Environments | Development, staging and production in every bin, with promote and a diff | Separate prefixes or buckets you set up | An environment resource for each, per application |
| How apps read config | A plain HTTPS GET with ETag and 304, from any HTTP client | A public GET through CloudFront, or signed S3 requests | The AWS SDK's session API, or the AppConfig Agent |
| After a change | The edge copy is purged automatically | A CloudFront invalidation, or wait for the cache to expire | Clients pick it up on their next poll |
| Gradual rollout | To a share of apps chosen by ?bucket=, raised, completed or aborted by hand | None | Deployment strategies that roll out over a set time, with bake time |
| Rollback | One click or hotcfg rollback, to any numbered version | Restore an earlier object version, then invalidate | Stop the deployment, or deploy an earlier version; automatic on a CloudWatch alarm |
| Types in your app | Generated from the schema with hotcfg types, in 10 languages | None | None |
| Access control | Keys limited to bins and environments (Pro), and team roles | IAM, bucket policies and origin access control | IAM policies |
| Change notifications | Webhooks to Slack, Discord, Teams, Google Chat, Mattermost, Telegram or your endpoint (Pro) | S3 event notifications to EventBridge, SNS or Lambda | Extensions that notify EventBridge, SNS or SQS on a deployment |
| Price | Free for 3 bins; Pro $19 a month per organization, unlimited reads | S3 storage and requests, CloudFront requests and transfer, invalidations past 1,000 paths a month | $0.0000002 per request and $0.0008 per configuration received |
What hotcfg doesn't do
- Automatic rollouts and rollback on an alarm. AppConfig can roll a deployment out over a set time and undo it when a CloudWatch alarm fires. A hotcfg rollout moves when a person moves it, and rollbacks are made by a person: one click, one CLI command or one API call.
- Custom validation code. hotcfg checks config with JSON Schema only; it doesn't run your own code against it.
- Private networking. hotcfg is read over the public internet. It can't be reached only from inside a VPC.
Moving from S3 or AppConfig
Download the current JSON file, create a bin with a schema that matches its shape, and upload it with hotcfg push. Point the app at the bin's URL with the fetching code for your language, and remove the bucket or the AppConfig resources once nothing reads them.
Facts about Amazon S3, CloudFront and AWS AppConfig are from AWS's documentation as of November 2026. Something out of date? Tell us.
Try it with your own config.
Free for 3 public configs (bins), each with development, staging and production. No card needed. Three short steps from sign-in to an app reading its config, in the dashboard or the CLI.