Managed App Hosting
Run APIs, full-stack apps, WebSockets, and workers as managed, long-running processes with health checks and logs.
ExploreAI sandboxes / Adios cloud platform
AI sandboxes.
For AI agents.
Give your AI agent a workspace to edit files, run commands, and preview the result. Connect databases, storage, and workflows, then deploy the application on Adios.
app.example.comHTTPSEdge routing & managed TLSTraffic to the promoted healthy release
Private service networking
Object storage & persistent volumes - 01 Apps & workers
- 02 Databases & storage
- 03 Domains & networking
- 04 Logs & workflows
Code execution for AI agents
How do AI agent sandboxes work?
An AI sandbox is a controlled environment where an AI agent can run code, work with files, and inspect results.
In Adios, the agent works in your project files. Review its edits and command output, open the running preview, and deploy a workload version when you are ready.
Connect your AI client through the Model Context Protocol (MCP). Authorize Adios with OAuth, then let your agent edit files, run commands, inspect logs, and open previews.
agent-workspaceInteractive example
src/app/page.jsx
export default function Page() { return ( <main> <h1>Hello, next idea.</h1> <p>Let's make it work.</p> </main> );}Make a change
The agent works in your project files.
SOURCE → EXECUTION → PREVIEWIllustrative data
The services behind your application
A full platform.
Around every AI sandbox.
Connect the app your agent builds to managed databases, storage, networking, and workflows. Keep runtime health and logs in the same platform.
Managed Databases
Add Postgres, pgvector, Redis, MongoDB, MySQL, or RabbitMQ and keep the dependency connected to the application.
ExploreObject storage & persistent volumes
Store uploads and assets in S3-compatible buckets, attach persistent volumes to workloads, and inspect available backups and build artifacts.
ExploreGlobal Edge Routing
Keep one public route while releases change. The gateway sends traffic to the promoted healthy version.
ExploreCustom Domains & TLS
Attach verified domains, managed TLS, and redirects to the application without rebuilding routing for every release.
ExploreWorkflows & Automation
Define scheduled jobs, webhooks, approvals, and ordered steps in YAML, then inspect the status and output of every run.
ExploreSecrets & Environment Variables
Keep required secret names in the manifest while their values stay outside Git and are resolved only where needed.
ExploreObservability & Logs
Connect build logs, runtime logs, versions, replicas, routes, and workflow runs to the source that produced them.
ExploreRuntime Security
Put the gateway in front, keep internal services non-routable, reference secrets outside source, and declare workload limits where supported.
ExploreFrom AI sandbox to production
Build, deploy,
and keep it running.
Keep build, start, health, region, and resource settings beside the source. Each release stays tied to its logs and health state.
Read the deployment guidePreview and production have separate lifecycles.
Test changes in a workspace preview. Deploy a workload version when it is ready, then verify the requested release and its routes.
01
Define the application
Build, start, health, resources, and region in adios.yaml.
02
Build a version
Keep the source, build artifact, and logs connected.
03
Check runtime health
Inspect the requested version and its replicas.
04
Route the release
Serve the promoted version through platform or custom domains.
Operate with context
Build logs and runtime logs from the CLI or dashboard
AI sandbox isolation and runtime controls
Control how it runs.
See what it uses.
Choose the runtime and resource limits, manage team access, and keep secrets outside your project source.
Explore adios.yaml ↗- Regions & replicas
- Select an available region and configure runtime replicas. Plan persistent data and connection limits as you add instances.
- Runtime resources
- Linux namespaces, cgroups, and resource limits help separate workloads where supported by the selected runtime.
- Private networking
- Connect managed databases, caches, and queues through generated private hostnames. Keep internal services off public routes.
- Team access
- Authenticated platform APIs scope access to the authorized user and team.
- Capacity
- Preview and resource allowances follow your plan. Check current pricing and limits ↗
A few useful answers
AI sandbox questions
From the Adios blogWhat is an AI sandbox?
Isolation, execution, and what your agent should be allowed to touch.
Read the guide ↗How do AI agents use an Adios sandbox?
An AI agent works in the workspace's project files, runs commands, and inspects their output. You can review file changes, run tests, and open a preview before deploying. Use the Adios workspace or connect an external AI client through MCP to work with the same project.
Can I connect my AI agent through MCP?
Yes. Connect a supported AI client to the hosted Adios MCP server and authorize access with OAuth. The client can work with workspace files, commands, logs, and previews. Review and approve deployments before execution.
How is code execution isolated?
Adios runtime controls include Linux namespaces, cgroups, and resource limits where the selected runtime supports them. Platform APIs scope access to the authorized user and team. Review the runtime, credentials, and network access available to a project before running agent-generated code.
What can I host on Adios?
Host web applications, APIs, WebSockets, and background workers as managed, long-running processes. Adios includes examples for Node.js, Python, Go, PHP, Ruby, .NET, and Next.js, with build commands, runtime configuration, health checks, and logs.
Can my application use databases, caches, and queues?
Yes. Add Postgres, pgvector, MySQL, MongoDB, Redis, or RabbitMQ as managed resources. Define them alongside the application in adios.yaml and use the generated private service hostnames and connection settings. Database, cache, and queue services run as their own workloads.
How do I keep files and application data?
Use S3-compatible buckets for objects and configured persistent volumes for filesystem data. Keep source in Git and review the backup records and recovery procedures available to your workload. Workspace source, runtime files, and database backups have different lifecycles.
Can I use my own domain?
Yes. Verify your domain, point DNS to the Adios-provided target, and attach it to your application. Adios manages TLS for verified domains and supports redirects. Generated platform routes are also available, and routing follows the promoted current release.
How do I deploy an app from an AI sandbox?
A workspace preview is a development runtime. Review the source and preview, then deploy a workload version through the dashboard, CLI, or an approved MCP workflow. The production release has its own build artifact, health state, logs, and routes.
How do regions, replicas, and resource limits work?
Choose a region available to your team and declare replica and resource settings in the manifest. Plan storage and database connections when adding replicas. Available capacity and preview allowances depend on your plan and runtime; check the pricing and deployment documentation for details.