The short answer
A Render Blueprint is a YAML file that declares interconnected services, databases, and environment groups. By default the file is named render.yaml and lives in the Git repository root. The name can be customized during setup. It is not portable. It is not an import format.
Use the file as an inventory. The migration guide says: “Render’s Blueprint specification can help you inventory repositories, commands, variable groups and disks. It is not an Unbroker import format.”
The source for the Render fields on this page is the render blueprint spec, read on 2026-11-16. This page does not quote Render prices. Railway vs Render is the side by side with Railway. A Railway file is mapped in railway.toml, field by field. A Heroku Procfile is mapped in Heroku Procfile, line by line. If you have already decided to move a service, start at Move a Render service to Unbroker. The comparison page is the Render page.
What a Blueprint declares
Root-level fields include services, databases, envVarGroups, and previews. Postgres databases are defined in the databases list, not in services. Do not define the same service, database, or group in more than one location.
A render.yaml example below is a set of field shapes from that spec. It is not a complete Blueprint and not one application. Command strings in the block are shapes, not documented defaults.
services:
- type: web
runtime: node
buildCommand: npm run build
startCommand: npm start
healthCheckPath: /health
preDeployCommand: npm run db:migrate
- type: web
runtime: static
buildCommand: npm run build
staticPublishPath: ./dist
- type: cron
runtime: node
schedule: "0 * * * *"
buildCommand: npm ci
startCommand: npm run job
envVarGroups:
- name: shared
envVars: []
previews:
generation: off
databases is a root-level list. This example does not invent the fields of a database entry. schedule is a cron expression. The expression in the block is a shape.
Service types and a destination per type
type is required. One of: web (a web service, or a static site when you also set runtime: static), pserv (a private service), worker (a background worker), cron (a cron job), keyvalue (Render Key Value; redis is a deprecated alias), workflow. You cannot modify type after creation.
The same spec’s glossary: a web service hosts a dynamic application at a public URL. A static site hosts HTML, CSS, and JS over a global CDN at a public URL. A render private service hosts a dynamic application that is not internet-reachable, for internal apps that only your other Render services can access. A background worker continuously runs code that does not receive incoming requests, and is described as ideal for processing jobs from a queue. A cron job executes a command or script on a predefined schedule. Render Key Value is fully managed, Redis-compatible storage.
Migrate from Render maps five rows. This page does not add rows.
| Render | What the migration guide says it becomes |
|---|---|
| Web service | Repository service with a matching preset or Dockerfile. |
| Static site | Static build preset and the correct output directory. |
| Background worker | Worker category with the consumer’s start command. |
| Persistent disk | Service volume, with an explicit file transfer. |
| Postgres | Managed database cluster or Apps PostgreSQL node. |
The same guide says: “Treat scheduled tasks and private-only services as separate design checks. Confirm the destination’s scheduling and connectivity behavior fits the workload before creating a public endpoint or substituting a long-running worker.”
keyvalue and workflow exist in the Blueprint. That table does not list them. This page does not invent a destination for either one, and it does not invent a Redis or workflow mapping.
Use Pro for workers, Dockerfiles, persistent volumes, and production domains. Worker services, Dockerfile builds, persistent volumes, and managed databases require Pro.
Build and start
runtime is required unless type is keyvalue or redis. Native runtimes are node, python, ruby, go, elixir, and rust. Special cases are docker (build an image from a Dockerfile), image (pull a prebuilt image from a registry), and static (static sites). You can change a service’s runtime after creation, except for static sites. runtime replaces the deprecated env field.
buildCommand is required for non-Docker-based services. It is the command Render runs to build the service.
startCommand is the command Render runs to start the service. It is required for most services. It is unsupported for static sites, which set staticPublishPath instead.
staticPublishPath is required for static sites. It is the directory that contains the static files to publish, relative to the repo root. Common examples in the spec are ./build and ./dist.
rootDir is the service’s root directory within its repo. Changes to files outside that directory do not trigger a build. If you omit it, Render uses the repo root.
dockerCommand is the command to run when starting a Docker-based service. If you omit it, Render uses the CMD defined in the Dockerfile. It is not supported for workflows.
dockerfilePath is the path to the Dockerfile, relative to the repo root. If you omit it, Render uses ./Dockerfile. It is not supported for workflows.
Dockerfile builds and prebuilt container images require Pro. A web service in the guide table becomes a repository service with a matching preset or Dockerfile. This page does not document a further mapping for runtime: image.
preDeployCommand and migrations
The render preDeployCommand is available for paid web services, private services, and background workers. If you set it, the command runs after the service’s buildCommand and before its startCommand. Render recommends it for database migrations and other pre-deploy tasks.
initialDeployHook is the other command. Render runs it after a service’s first successful deploy, once per service instance, including in preview environments. It applies to web services, private services, and background workers. afterFirstDeployCommand is a deprecated alias. For a command that should run before every deploy, use preDeployCommand instead.
This page does not document an Unbroker command sequence that replaces preDeployCommand. Migration commands require Pro. How a migration has to behave when you roll the code back is Database migrations on deploy.
Health checks, instances, and scaling
healthCheckPath is for web services only. It is the path of the service’s health check endpoint, and it always starts with a /.
numInstances, under Scaling, is the number of instances for a manually scaled service. If you omit it, Render uses 1 for a new service and retains the current value for an existing service. The value has no effect when autoscaling is enabled.
scaling is for an autoscaled service. minInstances and maxInstances are required. targetMemoryPercent or targetCPUPercent is optional. Each is optional if the other is set. Valid values are 1 to 90.
A plan field exists and sets compute resources. This page does not list plan SKUs or prices.
On Unbroker, Free runs one replica, up to 250m CPU and 256 MiB RAM. Advanced health checks require Pro. This page does not document an Unbroker equivalent for healthCheckPath, numInstances, or the scaling block. The replica note is Replica autoscaling.
Cron and workers
The render cron job schedule is schedule. It is required for cron jobs and omitted otherwise. The value is a cron expression. A cron job executes a command or script on a predefined schedule. A background worker continuously runs code that does not receive incoming requests.
Scheduled tasks are a separate design check. Confirm the destination’s scheduling behavior before you create a public endpoint or substitute a long-running worker.
This page does not document an Unbroker equivalent for that cron expression. Scheduled jobs and worker services require Pro. The difference between a process that stays up and a command that exits is Background workers and cron jobs.
envVars and envVarGroups
A Blueprint can create new environment variables or modify the values of existing ones. Render preserves existing environment variables even when you omit them from the file. Render does not support variable interpolation in Blueprint files.
A variable can be a hardcoded value. generateValue: true means that if the variable does not already exist, Render adds it and sets a randomized, base64-encoded, 256-bit value.
sync: false is prompted in the Render Dashboard during the initial Blueprint creation flow only. When you update an existing Blueprint, Render ignores environment variables with sync: false. Add new secret credentials to existing services manually. Render does not include sync: false variables in preview environments. You cannot apply sync: false to a variable defined in an environment group. Render ignores it. Those values do not carry. Load them by hand.
fromDatabase takes a name and a property, for example connectionString, and references a Render Postgres database. fromService takes a name, a type, and a property or envVarKey, and references another service. You can reference a service that is not defined in the Blueprint, but it must exist in the workspace or the Blueprint fails to sync. These referenced variables update to the current value of the referenced property on each Blueprint sync. They do not update immediately whenever that property changes. They do not carry. Replace them with destination references.
fromGroup adds every environment variable from an environment group.
render envVarGroups are the root-level groups. Each group has a name and a list of zero or more envVars. An environment group cannot reference values from your services or from other environment groups. You cannot define a variable with sync: false in an environment group.
previewValue can override environment variable values for web and private services in preview environments.
The migration guide says: “Recreate variables in the proper environment; values that refer to Render resources must be replaced with destination references.” How scope works once the values are ordinary environment variables is Environment variables and secrets.
Disks
disk attaches a persistent disk to a compatible service. name and mountPath are required. sizeGB defaults to 10. You can modify name and mountPath of an existing disk. You can increase sizeGB of an existing disk. You cannot reduce it.
The guide maps a persistent disk to a service volume, with an explicit file transfer. It also says: “File uploads on persistent disks require their own transfer; a database dump does not include them.” Persistent volumes require Pro. The volume note is Persistent storage volumes.
autoDeployTrigger, buildFilter, and previews
autoDeployTrigger on a Git-based service is commit (each commit to the linked branch, equivalent to the deprecated autoDeploy: true), checksPass (only if the linked branch’s CI checks pass), or off (auto-deploys disabled, equivalent to the deprecated autoDeploy: false). If both autoDeploy and autoDeployTrigger are present, autoDeployTrigger takes precedence. The field has no effect for a service that deploys a prebuilt Docker image. If you omit it, Render uses commit for a new service and retains the current value for an existing service.
buildFilter lists file paths, as glob syntax, that include or ignore changes when deciding whether to trigger an automatic build. Paths are always relative to the repo root. The example keys are paths and ignoredPaths. When synced, the value fully replaces an existing service’s build filter settings. If you omit the field on a service that already has build filter settings, Render replaces those settings with empty lists.
previews.generation is off, manual, or automatic. If you omit it, preview environments are disabled for any linked Blueprints. Setting the deprecated previewsEnabled: true is the same as automatic. This field does not affect configuration for individual service previews.
initialDeployHook is covered with migrations above. It runs once, after the first successful deploy.
This page does not document an Unbroker equivalent for autoDeployTrigger, buildFilter, initialDeployHook, or previews.generation. Pull request previews require Pro. The notes to read are Deploy on git push and Pull request preview environments.
render.yaml field by field
| render.yaml field | What it becomes |
|---|---|
services, databases, envVarGroups, previews | The root inventory. Postgres is in databases, not in services. Do not define the same service, database, or group twice. |
type: web | A repository service with a matching preset or Dockerfile. |
type: web and runtime: static | A static build preset and the correct output directory (staticPublishPath). |
type: pserv | A private service that is not internet-reachable. A separate design check. This page does not document a replacement type. |
type: worker | Worker category with the consumer’s start command. Worker services require Pro. |
type: cron and schedule | A cron expression for a command that runs on a schedule. A separate design check. This page does not document an Unbroker equivalent for the expression. Scheduled jobs require Pro. |
type: keyvalue (redis alias) and type: workflow | Blueprint types. The migration table does not list them. This page does not invent a destination. |
runtime | node, python, ruby, go, elixir, rust, docker, image, or static. Required except for keyvalue and redis. Static sites cannot change runtime later. |
buildCommand | The build command. Required for non-Docker services. |
startCommand | The start command. Required for most services. Unsupported on static sites. |
staticPublishPath | The static output directory, relative to the repo root. |
rootDir | The service root. Changes outside it do not trigger a build. Omitted means the repo root. |
dockerfilePath and dockerCommand | Dockerfile path (default ./Dockerfile) and the Docker start command (default Dockerfile CMD). Not supported for workflows. Dockerfile builds require Pro. |
runtime: image | Pull a prebuilt image. Prebuilt container images require Pro. This page does not document a further mapping. |
preDeployCommand | After build, before start, on paid web services, private services, and workers. Recommended for migrations. Migration commands require Pro. |
initialDeployHook | Once, after the first successful deploy, including in preview environments. This page does not document an Unbroker equivalent. |
healthCheckPath | Web services only. A path that starts with /. This page does not document an Unbroker equivalent. Advanced health checks require Pro. |
numInstances and scaling | Manual instance count, or min and max with an optional CPU or memory target from 1 to 90. This page does not document an Unbroker equivalent. Free runs one replica. |
plan | Sets compute resources. This page does not list SKUs or prices. |
disk | A service volume, with an explicit file transfer. sizeGB defaults to 10 and can increase, not shrink. Persistent volumes require Pro. |
autoDeployTrigger | commit, checksPass, or off. This page does not document an Unbroker equivalent. |
buildFilter | paths and ignoredPaths. Sync replaces the existing filter. Omitting it on a service that has a filter clears it to empty lists. This page does not document an Unbroker equivalent. |
previews.generation | off, manual, or automatic. Omitted means preview environments are disabled for linked Blueprints. This page does not document an Unbroker equivalent. Pull request previews require Pro. |
envVars value and generateValue | A hardcoded value, or a random 256-bit value when the variable does not exist yet. Recreate variables in the proper environment. Existing variables are kept when omitted from the file. |
sync: false | Prompted only on the initial create. Ignored on later syncs, omitted from preview environments, and ignored inside an environment group. Load the value by hand. |
fromDatabase, fromService, fromGroup | Render references. They do not carry. Replace them with destination references. |
previewValue | An override for web and private services in preview environments. |
| Variable interpolation | Not supported in Blueprint files. |
How this works on Unbroker
Migrate from Render says: “A Render migration begins with the running configuration, including any settings represented in render.yaml. Recreate the service behavior on Unbroker before moving production users or deleting source resources.”
The same guide’s translation is the five-row table above. It also says: “Once destination writes exist, returning DNS to the old database would lose those writes unless you reconcile them.”
Free is $0. It is one project, two services in total, and three deploy requests per UTC day. Unused deploys do not carry forward. They reset at 00:00 UTC. Each Free service runs one replica, up to 250m CPU and 256 MiB RAM, sleeps when idle, and wakes on the next request. Pro is $25 a month, including $25 of usage credit per billing cycle. The credit resets, does not roll over, and has no cash value. If eligible usage is under the credit, the monthly charge is still the plan price. Those terms are the plans page effective 23 Sep 2026, in USD and excluding tax, and they are on pricing and in Plans.
Dockerfile builds, prebuilt container images, worker services, persistent volumes, migration commands, pull request previews, scheduled jobs, one-off commands, advanced health checks, custom domains and managed TLS, and managed databases are not available on Free. They require Pro.
The field notes that already exist, and that this page does not restate as new product rules, are Database migrations on deploy, Background workers and cron jobs, Environment variables and secrets, Persistent storage volumes, Replica autoscaling, Deploy on git push, Pull request preview environments, and Docker Compose in production.
Checklist
- Treat
render.yamlas an inventory. It is not an import format. Read the running configuration, including settings the file represents, before you move production users or delete source resources. - List every root entry:
services,databases,envVarGroups, andpreviews. Postgres is indatabases. Do not count the same service, database, or group twice. - Record
typeandruntime. You cannot changetypelater. You cannot change a static site’s runtime later.keyvalueandworkflowhave no destination on this page. - Write down
buildCommand,startCommand,staticPublishPath,rootDir,dockerfilePath, anddockerCommand. - Move
preDeployCommandout as a migration step. It runs after build and before start.initialDeployHookruns once. Migration commands require Pro. - Copy
healthCheckPath,numInstances,scaling, andplan. This page does not document Unbroker equivalents for the check path, the instance count, or the scaling block, and it does not list plan prices. - Treat
scheduleandpservas design checks. Confirm scheduling and connectivity before you create a public endpoint or substitute a long-running worker. - Recreate variables in the proper environment. Load
sync: falsevalues by hand. ReplacefromDatabaseandfromServicewith destination references. Blueprint files do not interpolate variables. - Plan a file transfer for each
disk. A database dump does not include those files. You can increasesizeGB. You cannot reduce it. - Record
autoDeployTrigger,buildFilter, andpreviews.generation. This page does not document Unbroker equivalents for those fields. - Check the plan. Workers, Dockerfiles, persistent volumes, production domains, migration commands, scheduled jobs, pull request previews, one-off commands, advanced health checks, prebuilt images, custom domains, managed TLS, and managed databases require Pro.
- After the destination has new writes, returning DNS to the old database loses those writes unless you reconcile them.
FAQ
What is render.yaml for?
Every Render Blueprint is a YAML file that defines interconnected services, databases, and environment groups. By default the file is named render.yaml and lives in the Git repository root. The name can be customized during setup. Use it as an inventory of repositories, commands, variable groups, and disks. The source is the render blueprint spec.
Can I import a render.yaml on another platform?
No. A render blueprint is not portable, and it is not an Unbroker import format. The migration guide says the Blueprint specification can help you inventory repositories, commands, variable groups, and disks. Recreate the service behavior before moving production users or deleting source resources. railway.toml is not an import format either.
How does render preDeployCommand run migrations elsewhere?
preDeployCommand is available for paid web services, private services, and background workers. It runs after buildCommand and before startCommand. Render recommends it for database migrations and other pre-deploy tasks. For a command that should run before every deploy, use preDeployCommand, not initialDeployHook. This page does not document an Unbroker command sequence that replaces it. Migration commands require Pro. How a migration has to behave on a rollback is Database migrations on deploy.
What is a render private service (pserv), and what replaces it?
type pserv is a private service. It hosts a dynamic application that is not internet-reachable, for internal apps that only your other Render services can access. You cannot modify type after creation. Treat private-only services as a separate design check. Confirm the destination's connectivity behavior fits the workload before creating a public endpoint. This page does not document a second Unbroker type that replaces pserv.
What happens to sync:false and fromDatabase when you move?
sync:false is prompted in the Render Dashboard during the initial Blueprint creation flow only. On an update, Render ignores those variables, so add new secret credentials to existing services manually. Render does not include sync:false variables in preview environments, and it ignores sync:false on a variable in an environment group. Load those values by hand. fromDatabase (name plus a property such as connectionString) and fromService (name, type, and property or envVarKey) point at Render resources. They update on each Blueprint sync, not immediately when the property changes. Replace them with destination references, and recreate variables in the proper environment.
What replaces a render cron job schedule?
schedule is required for a cron job and omitted otherwise. The value is a cron expression. A cron job executes a command or script on a predefined schedule. Treat scheduled tasks as a separate design check, and confirm the destination's scheduling behavior before substituting a long-running worker. This page does not document an Unbroker equivalent for that cron expression. Scheduled jobs and worker services require Pro. The split between a process that stays up and a command that exits is Background workers and cron jobs.