# Ample vs Railway: which is better for deploying apps from coding agents?

Ample and Railway compared on databases, isolation and agent access, with sourced limits and steps to move a Railway service.

Checked November 5, 2026 · [Read as Markdown](/content/compare/railway.md)

## the short version

Railway is the better choice for services that need several regions, replicas or Docker images. Ample is the better choice when a coding agent should deploy alone: it signs up without a person, and each app runs in its own microVM with databases created automatically.

## Who this is for

Both platforms run real long-running processes from a repository, offer managed databases and work with coding agents. They differ in isolation, in how billing stops, and in how much an agent can do without a person.

## Ample and Railway side by side

| Round by round | Ample | Railway |
|----------------|-------|---------|
| Free plan | Yes. Free is $0 for up to 2 apps, with no card. | A 30-day trial with $5 of credit and no card; afterwards a Free plan with $1 of usage a month, 0.5 GB of memory per service and no custom domain. |
| Paid plans | Pro is $20 a month for up to 10 apps and Scale $99 a month for up to 50. Every plan includes compute, databases, S3 storage, a CDN and scheduled jobs. | Hobby $5 a month and Pro $20 a month per workspace, each including that much usage. Compute is billed per second: about $20 per vCPU and $10 per GB of memory a month. |
| Spending limit | A hard spending cap on every plan, on by default: apps pause and are never billed past it. | An optional hard usage limit (minimum $10) takes every workload offline when reached; services come back when the limit is raised. |
| How apps run | Each app runs as a normal long-running process in its own microVM with its own kernel. | Containers built from a repository or a Docker image, on Railway's own hardware. Isolation for regular services is not documented. |
| Long-running servers and workers | Yes. Web services and background workers (`--kind worker`) run as ordinary processes, with WebSockets, streaming responses and requests of 10 minutes or more. | Yes. Services run continuously; WebSockets are supported. |
| Languages | Node, Python, Go, Ruby, PHP, Java, .NET, Elixir, Rust and static sites, detected from the repository. | Node, Python, Go, PHP, Java, Ruby, Deno, Rust, Elixir and static sites through Railpack, or any Dockerfile. |
| Docker images | Not needed: Ample Computer builds from source and runs your start command; a Dockerfile's own steps are not run. | Yes. A Dockerfile is used automatically, and registry images can be deployed. |
| Databases | Managed Postgres (with pgvector), MySQL, Redis and MongoDB, each in its own microVM, created automatically when the code needs one. | One-click Postgres, MySQL, Redis and MongoDB templates, with HA variants; backups are turned on by the user. |
| Object storage | S3-compatible buckets, optionally published through Ample's CDN. | Private S3-compatible buckets; no public buckets. |
| Persistent disk | Persistent volumes attached to the app. | Volumes, one per service; services with a volume cannot use replicas. |
| Custom domains and TLS | Custom domains with automatic TLS. | Custom domains with automatic certificates; 2 on Hobby, 20 on Pro. |
| Regions | The USA and Europe, placed automatically; the USA is the default. | 4 regions: US West, US East, EU West, Southeast Asia. |
| Scheduled jobs | Scheduled jobs: a command or HTTP request on a cron schedule (any time zone) or interval, with retries and run history (`ample job create`). | Built-in cron services. |
| Agent access | CLI, remote MCP server (OAuth), A2A endpoint and agent skills. Deploys are synchronous and return JSON with an exit code: 0 live, 1 failed, 2 needs input. | Hosted MCP at mcp.railway.com plus a fuller local MCP, plugins for ChatGPT, Claude, Claude Code, Codex and Cursor, and an agent skill. |
| Agent signup without a person | Yes. `ample auth signup` creates an account with no browser, card or CAPTCHA; a person claims it by email within 48 hours. | Partly. `ssh railway.new` gives a temporary VM with no account (60 minutes to build, 24 hours to claim); `railway up` still needs a person to approve signup. |

Competitor facts come from that company's own pricing pages and documentation, checked on the date shown in the sources. Plans change; follow the source links for current numbers. Railway stops reading `railway.json` and `railway.toml` on December 1, 2026.

## Choose Ample when

- ✓ Each app should be isolated in its own microVM with its own kernel.  
- ✓ An agent should sign up, deploy and read structured failure diagnoses (who must act, why, and a suggested fix) without a person approving anything.  
- ✓ You want managed databases created automatically when the code needs one.

## Choose Railway when

- ✓ You need many regions, horizontal replicas, or a region outside the USA and Europe.  
- ✓ You deploy Docker images or rely on a Dockerfile build.  
- ✓ You want a template marketplace.

## Move a Railway service to Ample

1. **Write down build and start commands.** Railpack detects these implicitly. Copy them from the service settings, or from `railway.json` / `railway.toml`, which Railway stops reading on December 1, 2026.

2. **Resolve reference variables.** Replace `${{Postgres.DATABASE_URL}}` and `${{shared.X}}` with real values in a local env file, and drop `RAILWAY_*` variables, which do not exist elsewhere.

3. **Plan and deploy.** Ample detects the runtime and any database the code needs. Pass your env file; if you omit DATABASE_URL, Ample creates a managed database and injects it.
    
    ```
    $ ample plan --write && ample up
    ```

4. **Move data and schedules.** Dump Railway databases with `pg_dump`, `mysqldump` or `mongodump` and restore into the Ample database. Copy bucket objects with any S3 tool. Cron services become Ample jobs (`ample job create`).

5. **Point your domain.** Add the domain to the app, update DNS to the target Ample prints, and keep the old app running until the new one answers on the domain.
    
   ```
    $ ample domain add <app-name> www.example.com
    ```

## Frequently asked questions

**Does Ample have something like ssh railway.new?**  
The closest is `ample auth signup`, which gives an agent a real account (free-plan limits, 48 hours to claim) with no browser, card or CAPTCHA, and then deploys the repository it is working in. For a hands-on machine, `ample server create --template ubuntu-24.04` gives the agent its own VM: it runs commands with `ample server exec` and starts a supervised process with a public URL with `ample service start`.

**What happens when a deploy fails?**  
Ample returns the failure as JSON: who must act (the app, the account or Ample), a diagnosis, evidence from inside the app's VM and a suggested fix, so the agent can fix the code and redeploy without a person reading logs.

**Can I run a Docker image on Ample?**  
Ample Computer builds from source for its supported runtimes instead of running images or Dockerfile instructions; give it the start command. If your service only exists as an image, Railway is the better fit.

**Can Ample import my Railway project?**  
There is no one-click import. `ample plan --write` reads the repository and proposes the services and databases; the steps above cover variables and data.

## Sources · every claim about them, checked

- [Ample agent instructions (AGENTS.md)](/content/AGENTS.md) (checked 2026-10-14)
- [Railway pricing](https://railway.com/pricing.md) (checked 2026-10-14)
- [Railway free trial](https://docs.railway.com/pricing/free-trial.md) (checked 2026-10-14)
- [Railway plans](https://docs.railway.com/pricing/plans.md) (checked 2026-10-14)
- [Railway cost control](https://docs.railway.com/pricing/cost-control.md) (checked 2026-10-14)
- [Railway Serverless (app sleeping)](https://docs.railway.com/deployments/serverless.md) (checked 2026-10-14)
- [Railway Railpack builds](https://docs.railway.com/builds/railpack.md) (checked 2026-10-14)
- [Railway databases](https://docs.railway.com/databases.md) (checked 2026-10-14)
- [Railway storage buckets](https://docs.railway.com/storage-buckets.md) (checked 2026-10-14)
- [Railway volumes](https://docs.railway.com/volumes/reference.md) (checked 2026-10-14)
- [Railway regions](https://docs.railway.com/deployments/regions.md) (checked 2026-10-14)
- [Railway networking limits](https://docs.railway.com/networking/public-networking/specs-and-limits.md) (checked 2026-10-14)
- [Railway cron jobs](https://docs.railway.com/cron-jobs.md) (checked 2026-10-14)
- [Railway MCP server](https://docs.railway.com/ai/mcp-server.md) (checked 2026-10-14)
- [Railway plugins and connectors](https://docs.railway.com/ai/plugins-and-connectors.md) (checked 2026-10-14)
- [Railway free VM (ssh railway.new)](https://railway.com/free-vm.md) (checked 2026-10-14)
- [Railway for agents](https://railway.com/agents.md) (checked 2026-10-14)
- [Railway infrastructure as code](https://docs.railway.com/infrastructure-as-code.md) (checked 2026-10-14)
- [Railway reference variables](https://docs.railway.com/variables/reference.md) (checked 2026-10-14)

## Other comparisons

- [vs Render](/content/compare/render/index.html)
- [vs Fly.io](/content/compare/fly-io/index.html)
- [The best Railway alternatives in 2026](/content/compare/railway-alternatives/index.html)
