Source stays inspectable
The AI works on project files, so you can review a diff, test it, commit it, and keep the next change grounded in the same codebase.
GuidesAI deployment
AI app deployment
Deploy AI-generated code without losing the developer workflow.
AI can produce a useful first version, but developers still need source control, config, previews, logs, secrets, health checks, and a release path. Adios puts those pieces in one workspace-to-production flow.
Source to deploy
1AI request
2Workspace diff
3Preview runtime
4adios.yaml
5Production route
The problem
The hard part starts after the model writes code.
Copying generated files into a repository is only the beginning. A deployable app needs repeatable commands, environment boundaries, health behavior, logs, secrets, and a way to promote the version you actually checked.
Runtime stays explicit
Build, start, port, region, secrets, and dependencies belong in a manifest instead of being hidden in a chat transcript.
Deploy stays repeatable
The same app can be deployed from a template, a workspace, a local folder, or a Git-backed source path.
Workflow
A practical path from AI request to live app.
This is the workflow Adios is designed to keep in one place.
Start with source, not a transcript
Open a repository, local project, or Adios template. The AI agent edits files in a workspace that keeps project state, not just a copied answer.
Ask for a feature and keep the diff
Use AI to add the route, UI, worker, workflow, or database integration. Review the changed files, commands, and assumptions before anything goes live.
Run a preview with real logs
Build and run the app in a preview runtime. Check build output, runtime logs, health behavior, and the page or API response the change created.
Put the deploy contract in adios.yaml
Keep build commands, start commands, ports, secrets, managed resources, and routing choices beside the application code where they can be reviewed.
Deploy the verified version
Promote the current source into an Adios deployment, attach the generated route or custom domain, and continue from the same source-backed workflow next time.
Manifest
Put deployment behavior beside the code.
A small `adios.yaml` gives the app a reviewable runtime contract. The exact fields depend on the project, but the goal is the same: make the deploy path visible.
name: ai-assisted-api
runtime: node@24
region: de
build_cmd: npm ci && npm run build
start_cmd: node dist/server.js
health_cmd: node scripts/healthcheck.js
port: 8080
env:
DATABASE_URL: secret://DATABASE_URL
API_SIGNING_KEY: secret://API_SIGNING_KEYGood fit
Use Adios when...
- You want AI help but still need reviewable source, Git history, and repeatable deploys.
- Your app uses a normal framework such as Next.js, FastAPI, Django, Express, Go, Rails, or .NET.
- You need more than a static demo: APIs, workers, databases, queues, secrets, logs, or custom domains.
- You want one place to connect templates, workspaces, previews, runtime logs, and production routing.
Not the goal
Use a different path when...
- You only need a disposable prototype and do not plan to keep the source.
- You want to hand-manage cloud networking, VMs, proxies, and certificates directly.
- Your deployment cannot be described with explicit build and start commands.
FAQ
Common questions.
Is Adios only for AI-generated apps?
No. You can deploy an existing app, start from a template, or work in an AI-assisted workspace. The AI workflow is useful because Adios keeps generated code connected to source, preview, Git, and deployment.
What does adios.yaml do?
adios.yaml describes how Adios builds, runs, configures, and publishes the app. It keeps runtime decisions close to the code instead of only storing them in a dashboard.
Can I use my existing framework?
Yes. Adios has examples and templates for common web, API, and data-backed apps, including Next.js, Node.js, Python, Go, Ruby, .NET, Postgres, Redis, MongoDB, MySQL, and RabbitMQ.
Where does MCP fit?
MCP is one way for AI tools to work with platform context. The key Adios workflow is broader: source-backed edits, previews, logs, secrets, Git, and deploys stay connected even when the code started from an AI request.