# Connect an AI to your project

Configure MCP and an API key so an agent can read or edit project templates.

You can let AI inspect template structure, explain inputs, or edit a design in your project through **MCP (Beta)**. The connection exposes Tembuilder tools to your AI application; a project API key determines which operations it can perform.

## 1. Choose the connection for your task

Tembuilder provides two MCP connections:

| Connection | Purpose | Key required |
| --- | --- | --- |
| Docs MCP · `/docs/mcp` | Search and read public documentation. | No. |
| Project MCP · `/mcp` | Read or edit project templates and Assets. | Yes. |

For documentation questions, add a server at `https://<api-host>/docs/mcp` to your AI application without an API key. Use the API host shown on your project’s **MCP (Beta)** page. The steps below configure Project MCP.

<figure class="doc-sim data-assets-illustration workflow-illustration" aria-label="Docs MCP reads guidance · Project MCP works on templates">
<div class="sim-stage"><div class="ds-columns"><div class="sim-panel"><div class="ds-head"><strong>Docs MCP</strong><div class="wf-actions"></div></div><div class="wf-rows"><div class="sim-focus"><span>/docs/mcp</span><small>Public guides</small></div><div class=""><span>API key</span><small>Not required</small></div></div></div><div class="sim-panel"><div class="ds-head"><strong>Project MCP</strong><div class="wf-actions"></div></div><div class="wf-rows"><div class="sim-focus"><span>/mcp</span><small>Templates · Assets</small></div><div class=""><span>API key</span><small>Project-scoped</small></div></div></div></div></div>
<figcaption>Docs MCP reads guidance · Project MCP works on templates</figcaption>
</figure>

## 2. Create a project API key

In the intended project, open **API Keys** and create a key with the appropriate permissions:

- `mcp:read`: inspect templates, resources, and authoring context.
- Add `mcp:write`: allow creation, copying, and saving through writing tools.

The key determines which project AI can access. Put it in your AI application’s connection configuration; it does not need to appear in your editing prompt.

<figure class="doc-sim data-assets-illustration workflow-illustration" aria-label="API Keys → mcp:read · add mcp:write for editing">
<div class="sim-stage"><div class="sim-panel"><div class="ds-head"><strong>API Keys · MCP</strong><div class="wf-actions"></div></div><div class="wf-rows"><div class=""><span>mcp:read</span><small>Read content</small></div><div class="sim-focus"><span>mcp:write</span><small>Create and save edits</small></div></div></div></div>
<figcaption>API Keys → mcp:read · add mcp:write for editing</figcaption>
</figure>

## 3. Add the server to your AI application

Open the MCP or tool connections settings in your AI application. Use the endpoint shown on your project’s **MCP (Beta)** page and an `ApiKey` header. Example configuration:

```json
{
  "mcpServers": {
    "tembuilder-project": {
      "url": "https://<api-host>/mcp",
      "headers": { "ApiKey": "<YOUR_API_KEY>" }
    }
  }
}
```

This is a remote HTTP configuration example; the settings location and exact format depend on the client. The endpoint uses the API origin, rather than the dashboard URL. Clients using Authorization must send `Authorization: ApiKey ...`, rather than Bearer.

<figure class="doc-sim data-assets-illustration workflow-illustration" aria-label="MCP page → copy the endpoint and configure the connection">
<div class="sim-stage"><div class="sim-panel"><div class="ds-head"><strong>MCP · Project</strong><div class="wf-actions"></div></div><div class="ds-code sim-focus"><code>https://&lt;api-host&gt;/mcp</code><code>ApiKey: &lt;YOUR_API_KEY&gt;</code></div></div></div>
<figcaption>MCP page → copy the endpoint and configure the connection</figcaption>
</figure>

## 4. Start by reading a template

After connecting, try this request:

> Read the capabilities and list templates in the project. For the template I select, explain its structure and Parameters without changing content.

Once AI has found the intended template, describe the change you want and ask it to inspect a preview before saving. Then open the template in the editor to review its content and data.

Project MCP supports template creation, editing, and visual previews. To use the changes in a final PDF, [save and publish in the app](/docs/publishing/save-publish/), then [run a render](/docs/renders/preview-render/). MCP has no tool for publication or retained PDF jobs.

<figure class="doc-sim data-assets-illustration workflow-illustration" aria-label="workspace_capabilities → template_list → template_context">
<div class="sim-stage"><div class="sim-panel"><div class="ds-head"><strong>MCP · tools</strong><div class="wf-actions"></div></div><div class="wf-tools"><div class="wf-tool"><small>01</small><code>workspace_capabilities</code><span>Connection capabilities</span></div><div class="wf-tool"><small>02</small><code>template_list</code><span>Find project templates</span></div><div class="wf-tool"><small>03</small><code>template_context</code><span>Revision and template structure</span></div><div class="wf-tool sim-focus"><small>04</small><code>template_read</code><span>Relevant sources or schema</span></div></div></div></div>
<figcaption>workspace_capabilities → template_list → template_context</figcaption>
</figure>