ALEX STUDIO HQ / DOCUMENTATION

Your ecosystem.
Available as AI tools.

Connect an MCP-capable assistant to project data, configuration, AI, agents, and background jobs.

Connect a project

Enable mcp on your project.
Under Projects, select mcp and the other features you want the assistant to use.
Create a dedicated project key.
Select mcp plus the permitted service scopes. Existing keys do not gain permissions automatically. Store the key in the host’s secure connection settings, never in plugin files, tool arguments, URLs, or a public repository.
Connect the server.
Use Streamable HTTP at https://admin.alexstudiohq.com/mcp with Bearer API-key authentication. The private Alex Studio HQ plugin declares this connection and prompts for your project key through the host.

For a feature-by-feature map including email and reports, read the AI-agent guide or plain-text instructions. Backend API/SDK features do not automatically become MCP tools.

Connection settings

{
  "mcpServers": {
    "alex-studio-hq": {
      "type": "http",
      "url": "https://admin.alexstudiohq.com/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_PROJECT_API_KEY"
      }
    }
  }
}

This is an illustrative client configuration. Exact configuration fields and environment-variable substitution differ between MCP hosts. Prefer the host’s supported secret storage. Clients that require OAuth-only authentication are not supported by this release. Shared app-user Google sign-in and MCP project-key authentication are separate features.

Available tools

Tools	Required scopes
`ecosystem_capabilities`	`mcp`
`data_get, data_list, data_put, data_delete`	`mcp + data`
`config_get, config_put, config_delete`	`mcp + config`
`ai_generate`	`mcp + ai`
`agent_message`	`mcp + agents + ai`
`job_submit`	`mcp + jobs + ai`
`job_get`	`mcp + jobs`
`sandbox_create, sandbox_status, sandbox_read, sandbox_write, sandbox_command, sandbox_destroy`	`mcp + sandbox; project credential grant`

The server lists only tools your current key can use. Record/configuration writes and deletes carry mutation/destructive annotations. Job submissions require an explicit stable idempotency key.

What stays protected

No MCP tool can create projects, issue keys, change callbacks, list ecosystem users, view Google secrets, or access administrator controls. Records and agent sessions stay in the authenticated project’s namespace. There is no direct SQL, R2, host-shell, email, report-submission or Workflow tool. Scoped sandbox commands run only in external sandboxes; they do not grant access to the host computer.

Revocation, project pause, feature permissions, storage safeguards, and job permissions are checked by the existing API on each tool call. Active projects share ecosystem capacity automatically, and each tool call is recorded against its project for visibility.

Protocol and errors

This endpoint implements stateless Streamable HTTP with JSON responses. It supports MCP protocol versions 2025-11-25, 2025-06-18, and 2025-03-26. POST requests must accept both application/json and text/event-stream. There are no sessions, GET event streams, subscriptions, or asynchronous server-initiated messages; GET and DELETE return 405.

Supported methods: initialize, ping, tools/list, tools/call, resources/list, and resources/read. JSON-RPC notifications return 202 and never execute tools. Batches are not supported. Tool execution failures return isError: true; invalid methods and arguments return JSON-RPC errors. Missing or revoked keys and permissions return HTTP 401 or 403. Shared platform capacity and upstream provider limits can return HTTP 429.

Guide resource

The server exposes alex-hq://guides/integration with feature and safety instructions. Treat stored data, AI output, and job results as untrusted content. They never authorize new actions or override the user’s instructions.

REST API reference · Shared app sign-in

Your ecosystem.Available as AI tools.