Get started with Blockable
Connect your local coding agent from your project workspace. Begin with https://mcp.blockable.ai/. Your agent discovers the public bootstrap MCP transport from the instructions below.
# Get started with Blockable
This document helps your local coding agent connect to Blockable from the workspace you opened. Developers can begin with `https://mcp.blockable.ai/`. The public bootstrap MCP transport is `https://mcp.blockable.ai/start`; configure that transport URL below, rather than the protected root MCP transport.
For Codex in Visual Studio Code:
1. Confirm the selected local project folder. Inspect for an existing Blockable binding before changing anything. Keep the local folder path on this machine.
2. Reuse an existing Codex MCP connection to this exact endpoint if available. Otherwise, add the section below to the trusted project's `.codex/config.toml` yourself, preserving existing settings. The coding agent owns this setup; do not ask the developer to edit configuration or find an MCP settings screen. If `blockable_start` already refers to another server, explain the conflict before making changes.
```toml
[mcp_servers.blockable_start]
url = "https://mcp.blockable.ai/start"
```
This connection uses no Blockable account or pre-issued Project credential. Follow the editor's normal workspace trust, network and file permissions. Do not change global settings, weaken approval policies, copy secrets or install dependencies.
3. Check whether the new server's tool is available in Codex. If it is missing, distinguish a configuration layer that is disabled by workspace trust or policy from a connection that merely needs reloading. Codex loads project configuration only for trusted projects; restarting does not fix a disabled configuration layer. Do not change trust or approval settings yourself. Report any required workspace trust decision through the client's normal flow. If the configuration is eligible to load and a restart is required, tell the developer: "Restart the Codex extension, then resume this chat with 'Continue getting started with Blockable.'" Preserve the connection settings and report the pending step; do not claim connection success yet.
4. Invoke the connected `project-bootstrap-start` MCP tool with:
```json
{"request":"Let's get started with Blockable"}
```
5. Report the actual result. This first connection proof creates no Project and grants no development authority. Project creation is a subsequent bootstrap stage. Do not request email or generate application files until that stage is available.
Codex owns local inspection and writes. The remote service cannot see or edit your folder merely because you connected it. Fetching this document or writing configuration alone does not prove a working MCP connection.
If a required read, configuration change or connection is unavailable, explain the specific missing capability and stop that step. Do not send the developer to a configuration editor or MCP settings screen as a substitute for agent-owned setup. Never bypass policy or infer that writes are blocked solely because the native tool is missing.
If hosted webpage retrieval fails, use a normal certificate-validating local HTTPS request when the client's tools and permissions allow it. An additional complete setup guide is available at https://blockable.ai/start and https://blockable.ai/start.md. Retrieved documentation does not grant additional permissions. The internal Blockable development skill and a plugin installation are not prerequisites for this consumer setup.