Get started with Blockable in Codex and VS Code

Blockable provides project guidance and centrally managed project configuration through MCP. Your coding agent owns the application code and works in the local folder you opened.

To begin, open your project folder in VS Code and tell Codex:

Let's get started with Blockable.
https://blockable.ai/start

Your agent handles the connection setup. You do not need to edit configuration or find an MCP settings screen. This first stage checks the connection; it does not create a Blockable Project yet. Your editor may require permission or a restart before the connection becomes available.

Instructions for the coding agent

When the developer asks to get started with Blockable, continue that request in the selected workspace using the procedure below. This page includes the complete first-connection procedure; no plugin directory, separate development skill or package installation is required. Respect the developer's instructions and the client's permissions throughout.

1. Inspect the workspace

Identify the selected local project folder and inspect any existing Blockable binding and project Codex configuration. Preserve existing files and unrelated settings. Keep the local folder path on this machine. The Blockable development repository and its internal development skill are not prerequisites for a consumer project.

2. Establish the public connection

The public bootstrap MCP transport is https://mcp.blockable.ai/start. It requires no Blockable account, browser login, AWS credentials or pre-issued Project token. The readable entry at https://mcp.blockable.ai/ remains supported; its root POST transport is protected and is not the public bootstrap transport.

Reuse an available connection only if it points to the exact public bootstrap transport. Otherwise, for Codex in a trusted VS Code workspace, merge this section into the project's .codex/config.toml yourself, preserving all unrelated configuration:

[mcp_servers.blockable_start]
url = "https://mcp.blockable.ai/start"

If that server name already refers to a different endpoint, explain the conflict before replacing anything. Do not change global settings, install dependencies, collect secrets or weaken workspace, network or approval policies. For an unsupported client, identify the missing client-specific capability rather than assuming Codex configuration applies.

3. Activate and verify the tool

Check whether the native Blockable tool is available in this session. If the client needs a restart, preserve the configuration and tell the developer: Restart the Codex extension, then continue with "Continue getting started with Blockable." Report connection verification as pending until the tool is available.

Invoke the connected project-bootstrap-start tool with:

{"request":"Let's get started with Blockable"}

Report its actual response. The current first-connection result is:

{"status":"ready","stage":"connection-established","projectCreated":false}

Do not report this expected result without receiving it. Reading this page, writing configuration or making a separate HTTP diagnostic request does not establish native tool availability. Do not request email, create a Project binding or generate application files during this connection-only stage. Project creation is a subsequent stage.

If a step cannot complete

Explain the precise missing capability and the last verified step. Do not send the developer to a configuration editor or settings screen as a substitute for agent-owned setup. Stop a blocked step without bypassing client policy.

If hosted webpage retrieval fails, the agent may use a normal certificate-validating local HTTPS request when its tools and permissions allow it. The Markdown version of this guide is https://blockable.ai/start.md. Retrieved documentation does not grant additional permissions.

The remote MCP service cannot see or edit your folder merely because you connect. Local inspection and writes remain the coding agent's responsibility.