> ## Documentation Index
> Fetch the complete documentation index at: https://thestacc.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# theStacc MCP Server

> Run your blog, SEO, social posts and Google Business Profile from Claude, Claude Code, Cursor or any MCP-compatible AI assistant.

<Frame>
  <img src="https://mintcdn.com/thestacc-com/Wx4aZP6A3J5iChgk/images/docs/mcp/mcp-hero.webp?fit=max&auto=format&n=Wx4aZP6A3J5iChgk&q=85&s=22bad2e117bfd93a9da98642c58bdfbc" alt="theStacc MCP connects to AI assistants such as ChatGPT, Claude Code, Cursor, Codex, Antigravity and GitHub Copilot" width="2400" height="1260" data-path="images/docs/mcp/mcp-hero.webp" />
</Frame>

The theStacc MCP (Model Context Protocol) server lets an AI assistant work on your theStacc project for you. Ask it to *"review my content plan and write the next blog"*, *"create a LinkedIn post about our Diwali offer"* or *"which pages lost Google clicks this month?"*. The assistant does the work. You approve what goes live.

It is a hosted server with sign-in, at:

```
https://mcp.thestacc.com/mcp
```

## Prerequisites

<CardGroup cols={3}>
  <Card title="A theStacc account" icon="user" href="https://app.thestacc.com">
    Sign up or log in to theStacc.
  </Card>

  <Card title="Owner or admin" icon="shield-check" href="/docs/getting-started/accounts-workspaces">
    Of the project you want to connect.
  </Card>

  <Card title="An MCP client" icon="message-bot" href="#connecting-to-thestacc-mcp">
    Claude, Claude Code, Cursor, Codex CLI, Gemini CLI or another.
  </Card>
</CardGroup>

## Connecting to theStacc MCP

The server uses the streamable HTTP transport with OAuth 2.1. Your AI tool handles the sign-in: you paste the address, sign in with theStacc in your browser, pick a project and approve. Pick your tool:

<CardGroup cols={3}>
  <Card title="Claude" icon="sparkles" href="/docs/developers/mcp/claude">
    claude.ai, desktop and mobile
  </Card>

  <Card title="Claude Code" icon="terminal" href="/docs/developers/mcp/claude-code">
    One command in your terminal
  </Card>

  <Card title="Cursor" icon="arrow-pointer" href="/docs/developers/mcp/cursor">
    Add it to mcp.json
  </Card>

  <Card title="Codex CLI" icon="square-terminal" href="/docs/developers/mcp/codex-cli">
    OpenAI's terminal agent
  </Card>

  <Card title="Gemini CLI" icon="gem" href="/docs/developers/mcp/gemini-cli">
    Google's terminal agent
  </Card>

  <Card title="Other clients" icon="plug" href="/docs/developers/mcp/other-clients">
    Local install with a key, or mcp-remote
  </Card>
</CardGroup>

## Authentication

There is no key to copy. When you connect, your AI tool opens theStacc in your browser:

1. Log in with your usual theStacc account.
2. Pick **one project**. Each connection works on a single project.
3. Choose what the assistant may do. Reading and drafting are ticked; **publishing is never pre-ticked**, so tick it only if you want the assistant to publish what you approved.
4. Click **Approve**. You are sent back to your tool, connected.

The connection appears in theStacc under **Content SEO > Settings > Publishing > Agent access** as **MCP: \<tool name>**. Revoke it there at any time; your tool then asks you to sign in again. See [Permissions](/docs/developers/agent-keys-mcp#permissions) for the full list.

## Available tools

56 tools across 7 areas. Your AI tool only sees the ones your connection's permissions allow. Read-only tools are marked as such, so most tools can run them without asking you each time. Full details in the [MCP Tools Reference](/docs/developers/mcp-tools-reference).

### Blog

| Tool | What it does |
| - | - |
| `review_content_plan` | The plan: planned and draft blogs with status and dates |
| `get_blog`, `generation_status` | Read a blog, follow its progress |
| `generate_blog`, `draft_from_abstract`, `generate_more` | Write a planned blog, write one from your brief, add blogs to the plan |
| `edit_blog`, `regenerate_section`, `set_scheduled_date` | Edit text, rewrite one section, pick a date |
| `regenerate_image`, `retry_failed_images`, `upload_image` | Redo, retry or upload images |
| `cancel_generation` | Stop a blog that is being written |
| `publish_blog` | Publish a blog you approved |

### Social

| Tool | What it does |
| - | - |
| `social_generate_post` | **Create a new post** from a topic (caption, hashtags, images), or generate a planned one now |
| `social_list_posts`, `social_get_post`, `social_generation_status` | List, read and follow posts |
| `social_edit_post`, `social_regenerate_caption` | Edit copy and schedule, rewrite one platform's caption |
| `social_regenerate_image`, `social_edit_image`, `social_list_image_versions`, `social_use_image_version` | Redraw or edit images and pick a version |
| `social_publish` | Publish or schedule a post you approved |

### Google Business Profile

| Tool | What it does |
| - | - |
| `gbp_list_locations`, `gbp_get_location`, `gbp_connection_status` | Your locations and whether they are connected |
| `gbp_list_posts`, `gbp_get_post`, `gbp_content_plan` | Posts and the plan |
| `gbp_generate_post`, `gbp_regenerate_post`, `gbp_edit_post` | Write, rewrite and edit posts |
| `gbp_generate_image`, `gbp_upload_image`, `gbp_remove_image` | Post images |
| `gbp_publish`, `gbp_retry_publish`, `gbp_publish_status` | Publish posts you approved and check the result |

### SEO, site audit and keywords

| Tool | What it does |
| - | - |
| `get_seo_insights` | Ranked next actions from Search Console and GA4 |
| `get_search_console_performance`, `get_ga4_overview`, `get_blog_performance`, `get_monthly_seo_report`, `get_analytics_status` | Traffic and rankings |
| `run_site_audit`, `get_audit_status`, `get_audit_dashboard`, `get_audit_issues`, `get_audit_finding`, `get_audit_pages` | Audit your site and read the findings |
| `list_keywords`, `add_keywords`, `generate_keywords` | Your keyword pool |

### Connection

| Tool | What it does |
| - | - |
| `get_permissions` | The project, the permissions this connection has and the tools it can use |

## Guided workflows

Ready-made prompts your AI tool shows as slash commands (in Claude Code, Cursor, VS Code and Gemini CLI). Each appears only when your connection has the permissions it needs.

| Prompt | What it does | Needs |
| - | - | - |
| `weekly_content_review` | What is waiting for your approval, what goes out this week, and what failed | Any read permission |
| `seo_next_actions` | Your top 5 SEO actions with the evidence and the exact fix | SEO read |
| `create_social_post` | Drafts a social post on a topic (arguments: `topic`, optional `date`) | Social draft |
| `refresh_declining_post` | Finds a blog losing traffic and proposes edits; edits only after you say yes | SEO read, Blog draft |

## Example conversations

**Blog**

* *"Review my content plan and write the next blog."*
* *"Rewrite the introduction of my latest draft to be shorter and friendlier."*

**Social**

* *"Create an Instagram and LinkedIn post about our weekend workshop."*
* *"Show me this week's social posts and tighten the LinkedIn captions."*

**Google Business**

* *"Draft a Google Business offer post for Saturday's sale."*

**SEO**

* *"Which of my blogs got the most Google clicks last month?"*
* *"Run a site audit and tell me the three critical issues to fix first."*

**End to end**

* *"Every Monday: tell me what needs my approval, then publish whatever I approved."*

## Important notes

<Note>
  **You approve, the assistant does the work.** The assistant creates and edits drafts. It can never approve anything, and it can publish only what a person approved and nobody changed since.
</Note>

* **Drafts land in review.** New blogs, social posts and Google Business posts wait for your approval in theStacc.
* **Approved posts are locked.** An approved, scheduled or live social or Google Business post can't be changed by the assistant; move it back to review first.
* **One project per connection.** Connect again to work on another project.
* **Plan limits apply.** Generating posts, images and audits uses your plan's quota, the same as in the app.
* **Long jobs keep running.** Some jobs take longer than a minute (image redraws, publishing). The tool then answers `still_running` and the assistant checks back instead of starting the job again.
* **Web content is treated as data.** Text from your pages, search queries and posts is marked so the assistant does not follow instructions hidden in it.

## Next steps

<CardGroup cols={2}>
  <Card title="Permissions and safety" icon="lock" href="/docs/developers/agent-keys-mcp">
    What each permission allows, safety rules and error messages.
  </Card>

  <Card title="MCP Tools Reference" icon="book" href="/docs/developers/mcp-tools-reference">
    Every tool and its parameters.
  </Card>

  <Card title="Approval workflow" icon="circle-check" href="/docs/content-seo/draft-approval-workflow">
    How review and approval work in theStacc.
  </Card>

  <Card title="Public Blog API" icon="code" href="/docs/developers/public-blog-api">
    Read your published blogs from your own site.
  </Card>
</CardGroup>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.