Railway

Railway

This connector works with Railway through its public GraphQL API (https://backboard.railway.com/graphql/v2): projects with their environments and services, how each service runs, deployments with their build and runtime logs, the names of environment variables, and usage. It can redeploy or restart a deployment.

Setup (about 2 minutes)

1. In Railway open Account Settings, Tokens (https://railway.com/account/tokens) and create a token. Choose No workspace for an account token that reaches all your workspaces and can read your profile, or pick a workspace for a workspace token limited to that workspace (better for a team; railway_get_me does not work with it).
2. Paste the token into RAILWAY_API_TOKEN. It is sent as Authorization: Bearer <token>.
3. Check the setup with railway_list_projects.

Project tokens are not supported. A project token (project settings, Tokens) reaches one environment of one project and goes in a different header (Project-Access-Token); it cannot list projects. Use an account or workspace token.

Redeploys and restarts interrupt a running service: confirm first. Before railway_redeploy_deployment or railway_restart_deployment, tell the user which service and environment it affects and wait for their yes. A redeploy builds again and uses billed resources. Nothing here deletes or stops a service.

Secrets stay in Railway. railway_list_variable_names returns only the names of environment variables. railway_get_variable_values returns the values (passwords, keys, connection strings) and is installed switched off; turn it on only if the AI may see them. Runtime logs show whatever the app prints.

How to work

1. railway_list_projects (or railway_get_me for your workspaces), then railway_get_project for the environment and service ids and the latest deployment per service. railway_get_service_instance shows a service's settings in one environment.
2. railway_list_deployments (by service, environment or project) and railway_get_deployment. A FAILED deployment usually failed while building: railway_get_build_logs. A CRASHED one failed while running: railway_get_deployment_logs.
3. railway_redeploy_deployment or railway_restart_deployment on the latest deployment of the service.
4. railway_list_variable_names to check configuration; railway_get_estimated_usage and railway_get_usage for usage figures.

Usage units: CPU_USAGE is vCPU minutes, MEMORY_USAGE_GB and DISK_USAGE_GB (volumes) and BACKUP_USAGE_GB are GB minutes, NETWORK_TX_GB is egress in GB. With Railway's list prices at the time of writing (https://railway.com/pricing: CPU $20 per vCPU per month, RAM $10 per GB per month, volumes and backups $0.15 per GB per month, egress $0.05 per GB, a month counted as 43,200 minutes) the cost is for example CPU_USAGE x 20 / 43,200 dollars. Plan credits and discounts are not included.

Ids are UUIDs. Deployment status is one of SUCCESS, FAILED, CRASHED, BUILDING, DEPLOYING, INITIALIZING, QUEUED, WAITING, SLEEPING, REMOVED, REMOVING, SKIPPED, NEEDS_APPROVAL.

Paging: lists answer with pageInfo; while hasNextPage is true, pass endCursor as after.

Errors: Railway answers most problems with HTTP 200 and a message, which the connector reports as an error. Not Authorized: the token is wrong, revoked, or has no access to that project (or railway_get_me was called with a workspace token). Project not found, Deployment not found: wrong id or no access. HTTP 429: the hourly request limit of the plan is used up (100 on Free, 1,000 on Hobby, 10,000 on Pro); wait.

Other operations: railway_graphql_query and railway_graphql_mutation run any operation of the API. Railway publishes no SDL file, so railway_graphql_schema cannot load the schema; its introspection is public instead: run railway_graphql_query with query { __type(name: \"Deployment\") { fields { name args { name } } } } to look a type up. railway_graphql_mutation can change and delete anything the token can, so it is installed switched off (like railway_graphql_subscription); an admin can switch it on for a connector that should be able to do more.

Out of scope here: creating, changing or deleting projects, services, environments, variables, domains and volumes, rollbacks, and metrics charts.