Adios

Model Context Protocol

Use Adios from ChatGPT, Claude, and MCP-capable AI tools.

Adios exposes an OAuth-scoped MCP endpoint so AI clients can inspect workspaces, edit files, create previews, run commands, review code, and deploy approved changes through the same platform boundary as the web app.

Open Console

Public MCP config

Config
https://api.adios.dev/v1/mcp/config
Streamable HTTP
https://api.adios.dev/v1/mcp
SSE
https://api.adios.dev/v1/mcp/sse
Messages
https://api.adios.dev/v1/mcp/messages
Tools
https://api.adios.dev/v1/mcp/tools
transport: streamable_http, sse
scope: openid profile email mcp
resource: https://api.adios.dev/v1/mcp/sse

How it works

MCP turns an AI client into a scoped Adios operator.

The client does not get broad cloud credentials. It gets an OAuth-authorized MCP session with Adios tools that are scoped to the current user, team, workspace, and approval flow.

01

Connect client

Use the Adios MCP config endpoint.

02

Authorize

Complete OAuth with the mcp scope.

03

Use tools

Inspect source, run previews, and read logs.

04

Approve deploys

Require confirmation for sensitive actions.

Tools

Adios MCP functions.

Discovery and auth

Start by discovering platform capabilities and confirming the authenticated user and team context.

adios_capabilities

Describe Adios agent capabilities and tool groups.

auth_me

Return the current principal and team context.

Agent state and planning

Persist and inspect the plan behind an AI coding task so runs can be resumed and audited.

get_agent_state

Read session transcript, runs, plan steps, and events.

save_agent_plan

Persist the plan for the current coding task.

update_agent_plan_step

Mark a plan step passed, failed, blocked, or skipped.

append_agent_event

Append progress, tool, test, browser, approval, or deploy events.

Workspaces

Create, inspect, and prepare source-backed workspaces that AI tools can edit and run.

list_workspaces

List workspaces for a team.

get_workspace

Get one workspace by ID.

create_workspace

Create a workspace or project shell.

prepare_workspace_preview

Write files, generate adios.yaml when needed, build, start preview, and wait for a URL.

Files and manifests

Let an AI client inspect and change project files without losing source history.

list_workspace_files

List files in a workspace.

read_workspace_file

Read one file from a workspace.

write_workspace_file

Write one file to a workspace.

write_workspace_files

Write multiple files in one call.

write_workspace_adios_yaml

Create or replace adios.yaml.

move_workspace_file

Move or rename a file.

delete_workspace_file

Delete a file. Requires approval.

check_workspace_adios_yaml

Check manifest security and risky command patterns.

Commands, previews, and logs

Build, lint, start preview runtimes, poll long-running commands, and inspect logs.

build_workspace

Build an artifact from workspace contents.

build_workspace_dev

Queue the canonical workspace build command.

lint_workspace

Queue the canonical workspace lint command.

start_workspace_run

Start the preview runtime.

stop_workspace_run

Stop the preview runtime. Requires approval.

restart_workspace_run

Restart the preview runtime.

run_workspace_command

Queue lint, build, start, stop, restart, or run.

get_workspace_command_status

Get command status by command_id.

get_workspace_command_logs

Get command logs by command_id.

get_workspace_run_status

Get preview runtime status.

get_workspace_logs

Get runtime or build logs.

get_workspace_git_status

Get workspace Git status.

Review, deploy, and operate

Move from verified source to running workload, then inspect production logs.

security_review_workspace

Collect workspace context for security review.

deploy_workspace

Deploy the current workspace or build. Requires approval.

get_workload_logs

Get recent logs for a workload.

deploy_workload

Deploy a workload from an existing build. Requires approval.

Examples

Prompt patterns that map to MCP tools.

These are examples of what to ask an MCP-capable client after it is connected to Adios and authorized.

Adios MCP example console

Choose a prompt, review the MCP tool path, then ask your connected AI client to run it.

Create a workspace for my Next.js project...
01

Create a Next.js workspace

Create a workspace for my Next.js project, add the files for a minimal App Router app, generate adios.yaml, start a preview, and show me the preview URL.

View run

Simulated result

Adios creates the workspace, writes the starter files, adds the manifest, starts the preview runtime, and returns the running URL with command logs.

MCP tools used

auth_mecreate_workspacewrite_workspace_fileswrite_workspace_adios_yamlstart_workspace_runget_workspace_command_statusget_workspace_command_logs
02

Import and inspect a repository

Create a workspace from my Git repository, inspect the file tree, find the app entrypoint, and tell me what build and start commands Adios should use.

View run

Simulated result

Adios creates the workspace from the repository, lists source files, reads framework config, and proposes a deploy contract before writing anything.

MCP tools used

auth_mecreate_workspacelist_workspace_filesread_workspace_file
03

Fix a failed preview

My preview failed to start. Read the latest command logs, inspect package.json and adios.yaml, patch the issue, and restart the preview.

View run

Simulated result

Adios reads the logs, checks the manifest and package scripts, writes the smallest fix, restarts the runtime, and reports the new status.

MCP tools used

get_workspace_command_logsread_workspace_filewrite_workspace_filerestart_workspace_runget_workspace_run_status
04

Review before deploy

Review this workspace before deployment. Check adios.yaml for risky settings, run lint, summarize Git status, and tell me what still needs approval.

View run

Simulated result

Adios checks the manifest, queues lint, reads command logs, reports Git status, and separates safe fixes from approval-sensitive deploy actions.

MCP tools used

check_workspace_adios_yamllint_workspaceget_workspace_command_logsget_workspace_git_statussecurity_review_workspace
05

Deploy after I approve

Build this workspace, inspect the build logs, ask before deployment, then deploy the current version to the linked workload if I approve.

View run

Simulated result

Adios builds the workspace, streams logs, pauses for approval, deploys the approved build, and checks workload logs after release.

MCP tools used

build_workspace_devget_workspace_command_statusget_workspace_command_logsdeploy_workspaceget_workload_logs
06

Operate a live workload

Check the last 200 lines of logs for my workload, summarize errors, and suggest the workspace files I should inspect next.

View run

Simulated result

Adios reads production logs, groups errors, and points the AI client back to relevant source files for the next workspace change.

MCP tools used

get_workload_logslist_workspace_filesread_workspace_fileappend_agent_event

FAQ

MCP platform questions.

What MCP transport does Adios support?

The Adios MCP config advertises streamable HTTP and SSE. The public base URL is https://api.adios.dev/v1/mcp, with SSE at /sse and message posts at /messages.

How is Adios MCP authenticated?

Adios MCP uses OAuth. The advertised scope is openid profile email mcp, and tool calls are scoped to the authorized user and team context.

Can MCP deploy production workloads?

Yes, but deploy and destructive actions are modeled as approval-sensitive tools. The client should ask before calling approval-required tools such as deploy_workspace, deploy_workload, stop_workspace_run, or delete_workspace_file.

Is this the same as deploying my own MCP server?

No. This page covers using the Adios platform through its MCP endpoint. To build and host your own MCP server on Adios, use the MCP server deployment guide.