Web and desktop

Adios for ChatGPT.

Add Adios to ChatGPT as a custom connector and work with workspaces, previews, logs, and approval-gated deploys from a chat.

Install now Client docs

Connection: Custom connector
Authentication: OAuth
Access: Read, write, and approval-gated deploy

Setup

Install.

ChatGPT connects to the hosted Adios MCP endpoint as a custom connector. You sign in with Adios OAuth, and nothing is installed on your computer.

01
Turn on Developer mode
In ChatGPT settings, open Apps & Connectors, then Advanced settings, and turn on Developer mode. On Business and Enterprise workspaces, an admin may need to allow it.
02
Create the connector
Choose Create, name it Adios, paste the MCP server URL, and set authentication to OAuth.
Command
```
https://api.adios.dev/v1/mcp
```
03
Sign in
ChatGPT opens the Adios sign-in page. Approve access for the account you want ChatGPT to use.

Try a prompt

Start a new chat, choose Developer mode from the + menu, turn on Adios, then confirm the connection or start a real build.

Prompt 1

Use Adios MCP to confirm my user and team with auth_me.

Prompt 2

Build me a Next.js landing page and give me a working preview URL.

Prompt 3

Open this repository, fix the failed build, and show me the healthy preview URL.

Tool access

Capabilities.

The client uses the same scoped Adios tool surface. The live discovery endpoint remains the source of truth for schemas.

Workspaces

List, inspect, create, and prepare source-backed Adios workspaces.

list_workspacesget_workspacecreate_workspace

Source

Read and update project files while keeping changes visible in the workspace and Git state.

read_workspace_filewrite_workspace_fileget_workspace_git_status

Builds

Run lint and build commands, start a preview, and follow long-running command state.

lint_workspacebuild_workspace_devstart_workspace_run

Logs

Read command, build, preview, and production logs to diagnose real failures.

get_workspace_command_logsget_workspace_logsget_workload_logs

Deployment

Review the workspace and deploy only after the user explicitly approves the operation.

security_review_workspacedeploy_workspacedeploy_workload

View every MCP function

Full workflow

Build and verify.

“Can you add a pricing page to this Next.js app with monthly and yearly billing, a plan comparison table, and an FAQ? Match the existing design, make sure it works on mobile, and send me a preview link.”

01
Inspect
Open the Adios workspace, inspect its files, and identify the project commands.
create_workspacelist_workspace_filesread_workspace_file
02
Configure
Create or validate adios.yaml before running project commands.
write_workspace_adios_yamlcheck_workspace_adios_yaml
03
Fix and build
Run lint, make the smallest necessary source changes, and build the project.
lint_workspacewrite_workspace_filebuild_workspace_dev
04
Run and verify
Start the preview, inspect its logs and health, then return the verified URL.
start_workspace_runget_workspace_command_logsget_workspace_run_status

Help

FAQ.

Do I need to install anything?

No. ChatGPT connects to the hosted Adios MCP server directly. The Adios plugin with workflow skills is for Codex, Claude Code, Gemini CLI, and Kimi Code.

Which ChatGPT plans support custom connectors?

Custom connectors need Developer mode, which is available on paid ChatGPT plans. Free accounts can't add them.

How does authentication work?

ChatGPT registers with Adios OAuth and opens the Adios sign-in page. Approve access there instead of adding a static token.

Can this client deploy to production?

Yes. Deployment tools are available, but deployment and other sensitive operations remain approval-gated. The client should explain the intended action and wait for explicit confirmation.

Catalog