APIxStar Grow MCP

Website traffic, keyword and audience research in your AI conversation.

Connect your AI · 中文安装页

Install with one prompt

Paste this into a local ChatGPT Work / Codex desktop task:

Read https://grow.apixstar.com/chatgpt to install APIxStar Grow and sign me in. Check installation and authorization, briefly introduce what you can do, and guide me to start my first website research task in a new session.

For any session in the Claude Code desktop app:

/goal Read https://grow.apixstar.com/claude to install and use the APIxStar Grow plugin.

For WorkBuddy:

Read https://grow.apixstar.com/workbuddy to install APIxStar Grow and sign me in. Check installation and authorization, briefly introduce what you can do, and guide me to start my first website research task in a new session.

For a local task in TRAE desktop (formerly Trae Work):

Read https://grow.apixstar.com/trae-work to install APIxStar Grow and sign me in. Check installation and authorization, briefly introduce what you can do, and guide me to start my first website research task in a new session.

The guides are public: browsers get HTML and AI fetch tools get the same guide as Markdown, with no account, JavaScript or API key needed. The AI configures the client; you sign in with Google and approve access in the browser. A client with local execution is needed for automatic installation. Claude web can use a custom connector instead; local Claude Code Desktop follows the complete installation guide.

ChatGPT Work / Codex desktop

Use the CLI bundled with your desktop app when available, or a compatible Codex CLI. Check codex plugin --help first; older clients can use the direct MCP setup below.

codex plugin marketplace add apixstar/growthpilot-mcp
codex plugin add growthpilot@apixstar
codex plugin list --marketplace apixstar --json
codex mcp list --json

Find the APIxStar Grow server's exact name in the MCP listing, then run codex mcp login <name> --scopes tasks:read,tasks:write,billing:read. Plugin server names can include a namespace. Sign in and approve access in the browser. Start a new task to load the tools.

Claude Code

Use the complete ZIP linked from the Claude Code guide, extract it to a permanent folder, and install its local marketplace. The public Git marketplace may be older and lack the bundled skills; do not treat successful MCP registration as a complete installation.

claude plugin marketplace add "/absolute/path/to/growthpilot"
claude plugin install growthpilot@apixstar-download --scope user

Reuse a matching installation and valid authorization. The agent discovers the actual server name and starts claude mcp login <actual-server-name> in an interactive terminal, keeping it running for browser consent. If the client cannot run native CLI login, use /mcp → APIxStar Grow → Authenticate. Follow the guide for terminal support and targeted recovery; do not run concurrent login attempts or change permission modes.

After confirmed installation and authorization, introduce APIxStar Grow and guide the user to New session in Claude Code Desktop or a fresh terminal session. Report native verification as pending until that session loads both skills and the reference and calls growthpilot_account_get. A natural request naming APIxStar Grow can start the research; the installation goal ends at the handoff. Desktop MCP changes require a new session even where interactive terminal reload is supported.

WorkBuddy

Follow the complete WorkBuddy guide and download its versioned ZIP. Confirm the active MCP configuration through the running client's controls and the skill directory independently through its skill settings or bundled skill-creator. Inspect both installation scripts, then run:

node scripts/configure-workbuddy.mjs --server-url https://grow.apixstar.com/mcp --skills-dir "/absolute/path/to/confirmed/user/skills"

Add --config "/absolute/path/to/active/mcp.json" only when necessary. The helper backs up changed configuration, preserves unrelated entries, reuses a matching connection, and installs both complete skills plus references/source-selection.md. It does not authorize an account or read credential caches. configured_client_load_pending and client_oauth_pending mean client state still needs checking, not that login is necessarily missing. Use native Save, Trust and Connect only as the actual client state requires; reuse valid authorization and never put a fixed token in the configuration.

Check installation and authorization, briefly introduce the available capabilities, and guide the user to start their first website research task in a new session, even if the installation conversation already works. Preserve the project, endpoint, version, server name, confirmed skill paths, pending checks and any existing research goal. Keep the handoff short; no long technical verification prompt is required. In the new session, load both skills and the source-selection reference, discover available tools and call the real growthpilot_account_get before research. Do not run a paid query to verify setup; test token renewal only when requested.

TRAE / Trae Work

Follow the complete TRAE guide. Install both skills with the downloaded scripts/install-skills.mjs --skills-dir "/absolute/path/to/confirmed/user/skills". Confirm the directory from the running client independently of its MCP configuration. TRAE CN uses ~/.trae-cn/skills/ for global skills; other editions may differ. Without Node.js, copy both complete skill folders and preserve their references as described in the guide.

여기까지가 README 앞부분입니다. 나머지는 원문에서 읽을 수 있습니다. README 전문 보기 ↗