Blawgy MCP server

Run your blog from Claude, ChatGPT, Cursor or your own agent.

Blawgy researches keywords, writes SEO articles and publishes them to your site (WordPress, Shopify, Webflow, Wix, Framer, Ghost, Duda or your own API) on a schedule. This MCP server gives an AI agent everything the dashboard can do. Read your Search Console traffic, run keyword research, build and reorder the content plan, write and edit articles, draft and approve updates to live posts, change images, products, locations and settings, manage guardrails, webhooks and your team.

This repo holds the public listing (README, server.json, logo). The server itself runs inside Blawgy and is not open source.

Endpoint https://app.blawgy.com/mcp (Streamable HTTP) Auth OAuth 2.1 with PKCE and dynamic client registration, or an API key as a bearer token Plans every paid Blawgy plan, including clients of white-label partners Docs blawgy.com/mcp and the help article

Connect

Claude (web, desktop, mobile)

Customize, then Connectors, then Add custom connector. Name it Blawgy, paste the server URL, press Connect and sign in with your Blawgy account. Pick the sites and whether the connection can write. On Team and Enterprise plans an owner adds it under Organization settings. To use an API key instead, paste Authorization and Bearer <key> under Request headers.

Claude Code

claude mcp add --transport http blawgy https://app.blawgy.com/mcp
claude mcp login blawgy     # or run /mcp inside a session

With an API key, no sign-in needed

claude mcp add --transport http blawgy https://app.blawgy.com/mcp \
  --header "Authorization: Bearer blk_live_..."

Shared with your repo in .mcp.json (--scope project)

{
  "mcpServers": {
    "blawgy": {
      "type": "http",
      "url": "https://app.blawgy.com/mcp",
      "headers": { "Authorization": "Bearer ${BLAWGY_API_KEY}" }
    }
  }
}

ChatGPT

Settings, then Connectors, then Advanced settings, turn on Developer mode, press Create. Name it Blawgy, paste the server URL, choose OAuth, sign in. ChatGPT only supports OAuth, so keys do not work there. Developer mode on Plus and Pro is read only. Writing, publishing and settings changes need a Business, Enterprise or Edu workspace with Developer mode enabled by an admin.

Cursor

.cursor/mcp.json in a project, or ~/.cursor/mcp.json for every project. Leave out headers and Cursor prompts you to sign in.

{
  "mcpServers": {
    "blawgy": { "url": "https://app.blawgy.com/mcp" }
  }
}

With an API key

{
  "mcpServers": {
    "blawgy": {
      "url": "https://app.blawgy.com/mcp",
      "headers": { "Authorization": "Bearer ${env:BLAWGY_API_KEY}" }
    }
  }
}

VS Code and GitHub Copilot

.vscode/mcp.json, or run MCP: Add Server from the command palette.

{
  "inputs": [
    { "type": "promptString", "id": "blawgy-key", "description": "Blawgy API key", "password": true }
  ],
  "servers": {
    "blawgy": {
      "type": "http",
      "url": "https://app.blawgy.com/mcp",
      "headers": { "Authorization": "Bearer ${input:blawgy-key}" }
    }
  }
}

Drop inputs and headers to sign in with OAuth instead.

Windsurf (Devin Desktop)

~/.config/devin/mcp_config.json (Windows %APPDATA%\devin\mcp_config.json), or Cascade's menu, Open MCP config file.

{
  "mcpServers": {
    "blawgy": {
      "serverUrl": "https://app.blawgy.com/mcp",
      "headers": { "Authorization": "Bearer blk_live_..." }
    }
  }
}

Gemini CLI

gemini mcp add --transport http blawgy https://app.blawgy.com/mcp
/mcp auth blawgy

or with a key, gemini mcp add --transport http --header "Authorization: Bearer blk_live_..." blawgy https://app.blawgy.com/mcp.

Codex CLI

codex mcp add blawgy --url https://app.blawgy.com/mcp --bearer-token-env-var BLAWGY_API_KEY

or codex mcp add blawgy --url https://app.blawgy.com/mcp followed by codex mcp login blawgy for OAuth.

Gemini CLI extension

This repo is also a Gemini CLI extension (gemini-extension.json), so one command installs it with OAuth sign-in.

gemini extensions install https://github.com/adamgusky/blawgy-mcp

Cline

cline mcp install blawgy --transport http https://app.blawgy.com/mcp

Clients that only speak stdio can bridge with npx -y mcp-remote https://app.blawgy.com/mcp. See llms-install.md.

One-click links

• Cursor cursor://anysphere.cursor-deeplink/mcp/install?name=blawgy&config=eyJ1cmwiOiAiaHR0cHM6Ly9hcHAuYmxhd2d5LmNvbS9tY3AifQ==

• VS Code vscode:mcp/install?%7B%22name%22%3A%22blawgy%22%2C%22type%22%3A%22http%22%2C%22url%22%3A%22https%3A%2F%2Fapp.blawgy.com%2Fmcp%22%7D

Your own agent

Create a key in Settings, AI agents. Choose its reach (one site, all your sites, or every client site if you are a partner) and whether it can write. The key is shown once.

TypeScript SDK

import { Client, StreamableHTTPClientTransport } from '@modelcontextprotocol/client';

const client = new Client({ name: 'my-agent', version: '1.0.0' });
const transport = new StreamableHTTPClientTransport(new URL('https://app.blawgy.com/mcp'), {
  requestInit: { headers: { Authorization: `Bearer ${process.env.BLAWGY_API_KEY}` } },
});
await client.connect(transport);
const { tools } = await client.listTools();

Claude API (MCP connector, beta header anthropic-beta: mcp-client-2025-11-20)

"mcp_servers": [
  { "type": "url", "url": "https://app.blawgy.com/mcp", "name": "blawgy", "authorization_token": "blk_live_..." }
],
"tools": [{ "type": "mcp_toolset", "mcp_server_name": "blawgy" }]

OpenAI Agents SDK (Python)

MCPServerStreamableHttp(params={
    "url": "https://app.blawgy.com/mcp",
    "headers": {"Authorization": f"Bearer {os.environ['BLAWGY_API_KEY']}"},
})

Related MCP server: SEOPro AI MCP

What you can ask

• What grew and what slipped on my blog in the last 28 days, and why?

• Find low-difficulty keywords about soy candle care and add the best five to the plan.

• Write an article on fixing candle tunnelling and send it to WordPress as a draft.

• Which published posts are losing clicks? Draft an update for the worst one and show me the diff.

• Is yesterday's post indexed by Google and Bing yet? Recheck it.

• Add a guardrail that we never quote prices, and apply it to every site I run.

• Switch all my client sites to three posts a week and reschedule anything overdue.

Tools

Every site tool takes a site argument (the domain). The server's instructions tell the agent which sites the credential can reach. Slow work (research, writing, scans, syncs) returns a job id immediately and a status tool reports progress.

123 tools. "confirm" means the agent must confirm before the call runs. "Pro+" tools are hidden on lower plans.

Content plan

Articles

Article builder

Smart updates

Images and photo bank

Products

Location pages

Settings and integrations

Team, billing and help

Checkout, card entry, plan changes and account deletion stay in the browser. The relevant tools return a link for a human to open.

Security

• Scoped credentials. A key or OAuth connection reaches one site, all your sites, or (for partners) every approved client site. Each is read only or read and write. Read-only credentials never see write tools.

• Confirm step. Delete, publish, revert, disconnect and cancel require the agent to confirm with a one-line description of exactly what will happen.

• Same limits as the dashboard. Posts per week, credits and rate limits apply as they do when you click. Per credential, 120 calls a minute and 3000 a day. Agents never get admin access.

• Revoke any time. Settings, AI agents lists every key and connected app with its last use. Revoking takes effect on the next call.

• Stored as hashes. Keys and tokens are stored hashed. The key is shown once at creation.

• Logged. Every call records the tool, the site, the result and the time. Argument values are not stored.

• OAuth details. Authorization code with PKCE S256, dynamic client registration (RFC 7591), exact redirect URI match, 1 hour access tokens, rotating refresh tokens with reuse detection, resource indicator bound to the /mcp URL. Metadata at https://app.blawgy.com/.well-known/oauth-authorization-server and https://app.blawgy.com/.well-known/oauth-protected-resource.

Partners and agencies

A partner owner who connects reaches every approved client site from one connection, so a request like "switch every client to three posts a week" is a single call per site, driven by the agent. Clients can connect their own agents too, scoped to their own sites. Today the connection is made at app.blawgy.com, not on a white-label domain.

Links

• Product page blawgy.com/mcp

• Help article blawgy.com/help/connecting-ai-agents

• Model Context Protocol modelcontextprotocol.io

• Support adam@blawgy.com

License

The contents of this repository (README, server.json, logo) are MIT licensed. The Blawgy service and the MCP server behind app.blawgy.com/mcp are proprietary and governed by the Blawgy terms.