Adios

AI application environment

Real Workspaces for AI Agents.

Turn a prompt into source, run the work inside an isolated sandbox, keep it in a persistent workspace, verify it on a preview URL, and publish it behind a real domain.

A/

BrightBuild

/ workspace
main·agent run active
SourceAIPreviewLogs
brightbuild.run.adios.dev
updating

Project estimator

Price the work before it prices you.

A clear working estimate for independent teams and the clients who hire them.

Free estimateNo sign-upShareable

Estimate your project

Project

Marketing site

Timeline

4 weeks

ScopeMedium

Working estimate

€8,400

± 15%

Workspace preview

Inspect workspace source

0%
workspace preview · Frankfurtwaiting for checks

One workspace · one persisted agent plan · one preview · one release decision

One continuous project

The AI run ends. The application keeps going.

Adios connects the temporary work to durable project state. Source remains reviewable, sandbox output becomes evidence, and the accepted runtime gets a stable route.

Source-backed workspace

The work survives the run.

BrightBuild / main

Explorer

▾ src

▾ app

page.js

layout.js

▾ lib

estimate.js

◇ adios.yaml

◇ package.json

Git changes6

M src/app/page.js

A src/lib/estimate.js

src/lib/estimate.js

+31 −2

export function calculateEstimate(input) {

+ const base = rates[input.projectType];

+ const scope = multipliers[input.scope];

+ return roundToHundred(base * scope);

}

Agent event

Estimate logic added

Source, configuration, and the resulting diff remain attached to this workspace.

Sandboxed verification

Run the checks without losing the project.

3/3passed

Lint · 1.8s

Build · 4.2s

Health · 200

The execution boundary can be disposable. Its output returns to the source-backed workspace as evidence.

Release and route

Give the reviewed app a durable address.

CURRENT RELEASEhealthy

release_18

quote.brightbuild.co

TLS ready · generated route retained

Clear product boundaries

The sandbox is a step, not the destination.

Isolation is where the application proves it can run. The workspace is where its source, plan, configuration, and history live. The route is how real users reach the reviewed result.

Request

AI agent

Plans against the selected project context.

Source of truth

Persistent

Workspace

Source, Git, agent plan, configuration, and evidence.

Disposable boundary

Sandbox

Declared build, test, and preview tasks.

User-facing

Release + domain

Healthy version behind generated HTTPS or a custom host.

PromptReviewable sourceVerified runtimeLive app

Approval stays visible

Let the agent work. Keep the release decision.

An agent can inspect files, prepare changes, request declared tasks, and read the resulting preview and logs. You still see the evidence before a new version replaces the current one.

Source changesPreviewLogs
6 files

@@ estimator result @@

+ const estimate = calculateEstimate(input);

+ return <Estimate value={estimate} />;

health_path: /api/health

No secrets or platform credentials added to source.

Release evidence

Production buildpassed
Preview health200
Runtime errors0
Approve and deploy

Current release remains live until approval.

Before you connect an agent

Questions about the environment, answered.

What does the sandbox do?+

The sandbox is the isolated execution boundary used for declared builds, tests, and preview work. The source and resulting project state remain attached to the workspace instead of disappearing with one execution.

How does an AI agent access an Adios workspace?+

An MCP-capable AI client can authenticate through Adios and operate against a selected team and workspace using scoped project tools. It does not receive a general browser shell or unrestricted platform credentials.

Can an AI-created app use a real domain?+

Yes. Start with the generated HTTPS route, verify the application, and then attach a verified custom domain with managed TLS to the promoted current release.

Can I review the source before deploying?+

Yes. Workspace source, Git changes, declared task output, previews, logs, runtime configuration, and deployment controls stay visible so you can review the result before publishing it.

Prompt → workspace → running application

Build. Verify. Publish.