INFRASTRUCTURE

What it needs

• Compute: verified (Apps run in isolated x86_64 Firecracker microVMs that auto-pause when idle and wake on request; sizes are the priced VM sizes.)
• Postgres: verified (Managed PostgreSQL 16 runs in its own microVM and is auto-provisioned when an app needs a database and no DATABASE_URL is supplied.)

PREREQUISITES

Before you start

• Authorized access to the Heroku app source and its export tooling
• An inventory of every component in scope and out of scope
• A validated backup or copy before any cutover
• An Ample account token with servers:write, databases:read

TESTED CONFIGURATION

Exactly what was tested

• install: bundle config set --local path vendor/bundle && bundle config set --local without 'development test' && bundle install
• runtime: ruby
• size: s-1vcpu-1gb
• start: bundle exec rails server -b 0.0.0.0 -p $PORT
• template: ruby-3.4

STEP BY STEP

How to do it

1. 1

Inventory the source

List what Heroku app provides beyond the component you are moving. Out of scope here: Heroku add-ons other than Postgres, release-phase scheduling, dyno autoscaling and Heroku Scheduler are not migrated; use --release-command for release tasks.

2. 2

Source step 1

Export config vars with heroku config -s -a <app> into an env file kept out of the repository

3. 3

Source step 2

Translate the Procfile web process into the start command (--start or the package manifest's start script); worker processes need a separate --kind worker service

4. 4

Source step 3

Note buildpacks and add-ons in use; managed databases replace Heroku Postgres, other add-ons need their own plan

5. 5

Deploy on Ample in parallel

bundle install into vendor/bundle from Gemfile.lock (development and test groups skipped), then Puma via rails server reading PORT on the ruby-3.4 template with RAILS_ENV=production. Pass the exported variables with --env; keep the source running.

   ample deploy . --name <app-name> --public --env-file .env.production

6. 6

Validate before cutover

Run the app's own checks and, for data, compare counts and checksums; the example checks are the pattern self-tests (/p/configuration-secrets, /p/relational-records).

7. 7

Cut over

Add the custom domain with ample domain add, update DNS from the Heroku DNS target, keep the Heroku app until verified, then scale it down.

8. 8

Rollback

Point DNS or configuration back to the source. The source was never modified; deletion is a separate, user-executed step after validation.

EXAMPLES

Tested examples

• Rails pattern fixture (destination) (tests/deploy-canaries/rails-patterns): Verified destination basis: Rails compute.

SUCCESS CHECKS

How to know it worked

• destination app responds on its public URL (/ on the live URL, expect ample canary rails patterns)
• configuration-secrets check from the destination example (/p/configuration-secrets on the live URL, expect see the pattern fixture checks)
• relational-records check from the destination example (/p/relational-records on the live URL, expect see the pattern fixture checks)
• data or object counts and checksums match the source

LIMITATIONS

Know the limits

• Documentation only: nothing is executed automatically and no execution binding is offered.
• The source-side procedure is documented from Heroku app's standard tooling and was not executed in this catalog's evidence; the destination side was verified with the pattern fixture.
• No full source-product parity is claimed: Heroku add-ons other than Postgres, release-phase scheduling, dyno autoscaling and Heroku Scheduler are not migrated; use --release-command for release tasks.
• Verified on the ruby-3.4 template at s-1vcpu-1gb; region, compliance and request-duration limits are unknown.
• Managed PostgreSQL 16 only; extensions, connection limits and backup or restore procedures on the managed side are not verified beyond the pattern checks.

COST

Cost estimate

Estimated 10.00 USD per month (size prices from pricing.toml at build revision a1b8c38919e59cd035ebabaced73cf84ece24371).

• app server x1 s-1vcpu-1gb: 5.00 USD
• managed PostgreSQL database x1 s-1vcpu-1gb: 5.00 USD

Destination always-on monthly price of the tested sizes; apps auto-pause when idle. Source costs are unknown to Ample.

EVIDENCE

Verification evidence

• canary_run on 2026-09-20T21:55:28Z at revision 4f7cb9813c1f-dirty (CLI 4f7cb98): Destination side verified: the Rails pattern fixture deployed on Ample (bundle install into vendor/bundle from Gemfile.lock (development and test groups skipped), then Puma via rails server reading PORT on the ruby-3.4 template with RAILS_ENV=production) and its checks passed (the app built, started and answered on its public HTTPS URL with encrypted configuration delivered). The source-side export from Heroku app is documented from the vendor's standard tooling and was not executed by this catalog's evidence. (expires 2027-03-19T21:55:28Z)

EXECUTION

Execution binding

No execution binding. This recipe is documentation only; nothing is executed automatically.

NEXT ACTIONS

Typed next actions

• Browse the catalog index (GET /v1/catalog on the api origin; authentication not required, approval not required)
• Search published recipes by intent, stack and constraints (POST /v1/catalog/search on the api origin; authentication not required, approval not required)
• Prepare a side-effect-free deployment plan for an authorized project (POST /v1/catalog/plan on the api origin; authentication required, approval not required)
• Read the existing agent authentication setup (GET /mcp/setup on the api origin; authentication not required, approval not required)
• Browse Heroku app (GET /v1/catalog/nodes/migration%3Aheroku-app on the api origin; authentication not required, approval not required)
• Browse Rails (GET /v1/catalog/nodes/stack%3Aframework-rails on the api origin; authentication not required, approval not required)
• Browse Migrate compute (GET /v1/catalog/nodes/intent%3Amigrate-compute on the api origin; authentication not required, approval not required)

Actions describe possible next steps. They are typed data, not commands, and grant no permission. Public discovery never provisions anything; planning requires your own authenticated token and approval happens in your client.