Connect your assistant

Search live sources, follow results and sharpen a draft answer before it is given.

Connect from a chat app

Use a custom connector with this server URL, then sign in to AI Power Ups and allow the connection. No API key or local MCP package is needed. Manage or disconnect it from Connected apps. Public store listings are not yet available.

Server URL
https://api.powerups-ai.store/mcp
claude.ai, Claude Desktop and mobile

On Claude web or Desktop: Customize → Connectors → + → Add custom connector. Paste this server URL, then connect and sign in. Team and Enterprise owners must add the connector for their organization first. Once connected, you can also use it on mobile.

Tested 2026-09-19 — Claude web (OAuth). Tool calls and follow-ups. Host removal and fresh reconnection still need separate testing.

Documented setup — Claude Desktop and mobile (OAuth). We have not completed our own end-to-end test in these native apps.

Official setup guide →
ChatGPT

Enable Developer mode in Settings → Security and login, then open chatgpt.com/plugins and select + to add an app. Paste this MCP server URL, choose OAuth, then sign in and allow the connection. Availability depends on your account and workspace policy.

Tested 2026-09-19 — ChatGPT web custom connection (OAuth). Tool calls and follow-ups. In our 19 September test, removing the connection in ChatGPT did not revoke the server grant; revoke from Connected apps in AI Power Ups to be sure.

Documented setup — ChatGPT native mobile (OAuth). We have not completed our own end-to-end test. A custom connection does not establish directory availability.

Official setup guide →
Gemini app

Set up on the web at gemini.google.com → Settings → Connected Apps (or Personal Intelligence) → Custom Apps → paste this MCP URL, then sign in. Google currently requires a personal account, age 18+, US availability, English and Keep Activity enabled; work and school accounts are not supported. Once connected on the web, you can also use it on mobile.

Documented setup — Gemini web and mobile (OAuth). We have not completed our own end-to-end test with an eligible account.

Official setup guide →

Grok

Grok reaches AI Power Ups through the paths below. All of them talk to the same server, https://api.powerups-ai.store/mcp. In the Grok app you sign in through your browser and can disconnect at any time from Connected apps.

Grok app (grok.com, iOS and Android)

grok.com/connectors → New Connector → Custom → paste this server URL, then sign in to AI Power Ups and allow the connection. On mobile: Settings → Connectors. Grok Business and Enterprise admins add the same URL as an "Other" custom MCP connector in console.x.ai first.

Server URL
https://api.powerups-ai.store/mcp

Grok Build

Run once in a terminal. Leave out the --header flag to sign in through your browser instead; Grok Build keeps the OAuth token itself.

Terminal
grok mcp add --transport http ai-power-ups https://api.powerups-ai.store/mcp --header "Authorization: Bearer <YOUR_API_KEY>"

Paste API keys only into tools you run yourself, never into a chat window. Keys are revocable from Connected apps. Grok Build can also install the AI Power Ups plugin, which signs in through your browser: grok plugin install https://github.com/RoeySmallTree/ai-power-apps --trust

Connect manually

Create an API key in your dashboard for the API-key examples below. Claude Code and the published plugins also support browser sign-in. Replace the placeholder with your key and keep it private.

Claude Code

Run once in a terminal. --scope user makes this server available across your projects. Replace the placeholder with your API key. To sign in through your browser instead, omit --header and use /mcp to authenticate.

Documented setup — Claude Code manual configuration (API key or OAuth). Configuration checked; end-to-end acceptance of these manual routes remains pending. The plugin is tested separately.

Official setup guide →
Terminal
claude mcp add --scope user --transport http ai-power-ups https://api.powerups-ai.store/mcp --header "Authorization: Bearer <YOUR_API_KEY>"

Cursor

Merge the mcpServers entry into .cursor/mcp.json in a project or ~/.cursor/mcp.json for all projects. Preserve any existing servers. Open Connect for an install button filled with your key.

Documented setup — Cursor manual and install-link setup (API key). We have checked the configuration format, but have not completed our own end-to-end test in Cursor.

Official setup guide →
JSON
{
  "mcpServers": {
    "ai-power-ups": {
      "url": "https://api.powerups-ai.store/mcp",
      "headers": {
        "Authorization": "Bearer <YOUR_API_KEY>"
      }
    }
  }
}

VS Code

Merge into .vscode/mcp.json, preserving existing servers. Open the Command Palette → MCP: List Servers, select ai-power-ups and start it. Open Connect for an install button filled with your key.

Documented setup — VS Code manual and install-link setup (API key). We have checked the configuration format, but have not completed our own end-to-end test in VS Code.

Official setup guide →
JSON
{
  "servers": {
    "ai-power-ups": {
      "type": "http",
      "url": "https://api.powerups-ai.store/mcp",
      "headers": {
        "Authorization": "Bearer <YOUR_API_KEY>"
      }
    }
  }
}

Gemini CLI

Gemini CLI requires Gemini Code Assist Standard/Enterprise or a paid Gemini or Enterprise Agent Platform API key for model access. Personal Google sign-in is no longer supported. This is separate from your AI Power Ups connection. Merge the mcpServers entry into ~/.gemini/settings.json without replacing other settings. Restart Gemini CLI, then use /mcp to check the connection. Google's access notice

Documented setup — Gemini CLI manual setup (API key). We have not completed our own end-to-end test with supported model access.

Official setup guide →
JSON
{
  "mcpServers": {
    "ai-power-ups": {
      "httpUrl": "https://api.powerups-ai.store/mcp",
      "headers": {
        "Authorization": "Bearer <YOUR_API_KEY>"
      }
    }
  }
}

REST API

Any HTTP client works. Start with the catalog, then inspect a capability.

Terminal
curl -H "Authorization: Bearer <YOUR_API_KEY>" https://api.powerups-ai.store/v1/catalog

How your assistant uses it

  1. Discover: list capabilities and see which are available.
  2. Inspect: read the exact input fields and examples for a capability.
  3. Execute: run a search and receive records with reusable identifiers.
  4. Follow up: request more results, adjust the search or open a record's details.

For example, search academic papers with academic.search, then use the returned search_id with follow_up and action: "more".

Performance stages

Sharpen puts more thought behind an answer before your assistant gives it. Your assistant hands over the task and its draft, and gets back what it needs to make the answer sharper. Each stage goes further than the last.

Stock means Sharpen is off. Enable it in your settings and choose Stage 1, Stage 2, Stage 3 or Stage 3+. Ask “take it to Stage 3+” to sharpen a single answer at that stage; the assistant passes stage: "stage3+" to sharpen. Omit stage to use your saved setting. A one-answer choice does not change that setting, and usage reflects the work that actually completes.

  • Stage 1: A first edge. Loose ends pulled tight, the answer comes back cleaner. Typically 5 usage tokens per sharpen.
  • Stage 2: Sharper still. What slipped past comes into focus. Typically 30 usage tokens per sharpen.
  • Stage 3: Deep sharpening for the answers that carry weight. Typically 120 usage tokens per sharpen.
  • Stage 3+: Everything Sharpen has, at once. For when it has to be right. Typically 240 usage tokens per sharpen.

The task, context and draft are processed by third-party AI providers. Share only content you have permission to send. Read the privacy policy for retention and providers.

Projects

Your assistants also get a private file repository they share: ask one to “save this to a project”, and another, even from a different vendor, can pick it up when you say “we're working on the plant bot project”. Files stay in your account until you delete them, and you can browse, upload and download them under Projects in the dashboard. See the Projects reference for the tools and limits.

Usage tokens

Free sources cost no tokens. Paid searches and Sharpen use tokens based on provider cost, rounded up per call. The free plan includes 500 tokens each month. See plan allowances and pricing.

Integration guide for developers

Building an application rather than connecting an assistant? Start with the developer quickstart and the capability reference. Agents can read the integration guide or the OpenAPI document for schemas, examples and error handling.

Support

Email roey@smalltree.io. Include the capability and time of the problem; never send passwords or API keys.