Build with AI in a real development workspace.

Give the Adios agent the same project context your team uses: source files, Git state, declared tasks, preview output, logs, configuration, and the path to deployment. The work stays visible while you decide what to keep.

Start building Read the workspace docs

development workspaceAdios

Project source

▾ src

▾ app

page.tsx

layout.tsx

▸ api

◇ adios.yaml

◇ package.json

◇ .gitignore

Git changes2

M src/app/page.tsx

A src/app/Status.tsx

src/app/page.tsx+18 −6

18export function ProjectStatus() {

19+ const release = useRelease();

20+ return <Status state={release.state} />;

21}

Preview

app.run.adios.dev

Latest log

GET /health 200 · 18ms

Adios agent

Developer guided

Review

Add release health to the project dashboard and verify the preview.

Plan · 4 steps

✓Inspect source

✓Update component

✓Run build

4Check preview

Review changes

Context stays with the project

The agent works from project files and configuration instead of a pasted summary that goes stale.

Changes stay inspectable

Review the affected files, Git diff, task output, preview, and logs before accepting the result.

Release remains a decision

AI can help prepare and verify a change without turning every generated edit into a deployment.

Where AI workflows break

Stop rebuilding project context inside every AI conversation.

A chat window can suggest code, but it does not automatically know the current repository, runtime configuration, failed build, preview state, or deployment history. Developers lose time moving that context between tools and then proving the result still works.

What changes with Adios

Put AI where the source and verification evidence already live.

An Adios workspace keeps AI-assisted edits beside the project files, Git state, declared tasks, previews, and logs. The developer stays in control of the diff and the release decision.

The working loop

Move from request to reviewed change without losing the thread.

The workspace carries the evidence from the first instruction through the final review.

Step 01
Start with owned source
Open the repository or project that will own the resulting code.
Step 02
Let AI inspect before editing
Use the current tree, relevant files, and project configuration to form the plan.
Step 03
Run the project’s checks
Use declared build and verification tasks, then inspect their output and runtime logs.
Step 04
Review, commit, and deploy
Keep the diff visible and choose when the accepted version moves forward.

Review evidence

You can see what the AI changed and how the project responded.

The useful output is not just a generated answer. It is a reviewable project state with verification evidence.

Files and Git diff

Inspect the concrete source changes instead of accepting an opaque generated bundle.

Declared task output

Follow build and project-check output without exposing a general browser shell.

Preview and logs

Open the candidate application and read the logs tied to that workspace state.

Deployment context

Keep source, runtime configuration, health, and release actions connected.

Questions before connecting AI

Know the working boundary before the first change.

Is this a replacement for my local editor?

It does not have to be. You can work in the Adios web workspace or keep a local project connected through the Adios CLI, then use the same source-backed preview and deployment path.

Can I review AI-generated changes before they are deployed?

Yes. AI-assisted work lands in project files where you can inspect the diff, run declared checks, open a preview, and review logs before choosing what to commit or deploy.

Does the browser workspace expose a terminal shell?

No. The browser workspace uses scoped project operations and declared tasks. A general interactive shell is not part of the current browser experience.

Can I start with an existing repository?

Yes. A workspace can start from existing source so the AI works against the project your team already owns rather than an isolated code snippet.

Start with real source

Bring AI into the project without giving up review.

Open real source, ask for a change, verify the result, and keep the accepted work connected to Git and deployment.

Start building