AI assistants (MCP)

Skavelo has an MCP server, so an AI assistant can work with your workspace for you. Ask it to write a walkthrough, draft a playbook from a support macro, add a lookup skill from your API docs, or go through Needs attention and suggest fixes.

The server's address is:

https://api.skavelo.com/mcp

You'll also find it on the AI assistants page in the dashboard sidebar, with these steps for each assistant.

Connect an assistantCopy Anchor Link

Menu names can differ a little between versions of each app. In every case, the last step is the same: Skavelo opens in your browser, you sign in if asked, choose the workspace and what the assistant can do, and click Connect.

ClaudeCopy Anchor Link

On the web and in the desktop app:

1. Open Settings, then Connectors.
2. Click Add custom connector. Name it Skavelo and paste the address above.
3. Click Add, then Connect next to Skavelo.
4. Approve the connection in Skavelo.
5. In a chat, turn Skavelo on from the tools menu and ask, for example, "List my Skavelo agents."

On Claude Team and Enterprise plans, an owner may need to add the connector for your organization first.

ChatGPTCopy Anchor Link

1. Open Settings, then Apps and connectors.
2. If there's no option to add your own, open Advanced settings and turn on Developer mode.
3. Create a connector. Name it Skavelo, choose OAuth for authentication, and paste the address above.
4. Approve the connection in Skavelo.
5. In a new chat, add Skavelo from the tools menu and ask, for example, "List my Skavelo agents."

Which ChatGPT plans can add connectors, and whether a connector can make changes, is set by OpenAI and can change. If these steps don't match what you see, check OpenAI's help for custom connectors.

Claude CodeCopy Anchor Link

claude mcp add --transport http skavelo https://api.skavelo.com/mcp

Start Claude Code, run /mcp, choose skavelo, then Authenticate, and approve the connection in Skavelo. Add -s user to the command to use Skavelo in every project.

CursorCopy Anchor Link

Add Skavelo to Cursor: Cursor opens and asks to install it; click Install.

Or add it yourself: open Cursor Settings, then MCP, and add a server, or add this to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "skavelo": { "url": "https://api.skavelo.com/mcp" }
  }
}

Then, in Cursor's MCP settings, click Needs login next to skavelo and approve the connection in Skavelo.

VS CodeCopy Anchor Link

Run MCP: Add Server from the command palette, choose HTTP and paste the address, or add this to .vscode/mcp.json:

{
  "servers": {
    "skavelo": { "type": "http", "url": "https://api.skavelo.com/mcp" }
  }
}

Start the server when VS Code asks, allow it to sign in, and approve the connection in Skavelo. Then use Skavelo from Copilot Chat in Agent mode.

To add Skavelo's skills as well, run Chat: Install Plugin from Source and enter https://github.com/Skavelo/ai-assistants. It adds the skills and the connection together.

Gemini CLICopy Anchor Link

gemini extensions install https://github.com/Skavelo/ai-assistants

The extension adds Skavelo's skills and the connection. Start Gemini CLI, run /mcp auth skavelo and approve the connection in Skavelo.

AntigravityCopy Anchor Link

In the Antigravity app or IDE, add this to ~/.gemini/config/mcp_config.json:

{
  "mcpServers": {
    "skavelo": { "serverUrl": "https://api.skavelo.com/mcp" }
  }
}

Open Settings, then Customizations, click Refresh under Installed MCP Servers, then Authenticate next to skavelo. Approve the connection in Skavelo; Antigravity then shows a code to paste back into it.

In the Antigravity CLI, agy plugin install https://github.com/Skavelo/ai-assistants adds Skavelo's skills and the connection together. Sign in from /mcp.

CodexCopy Anchor Link

The Codex CLI, the Codex IDE extension and the ChatGPT desktop app share one setup. In your terminal:

codex mcp add skavelo --url https://api.skavelo.com/mcp
codex mcp login skavelo

Approve the connection in Skavelo. In the ChatGPT desktop app, you can instead open Settings, then MCP servers, choose Add server, pick Streamable HTTP and paste the address. Save, Restart, then Authenticate.

WindsurfCopy Anchor Link

Windsurf is now Devin Desktop. In Cascade's MCP settings, choose Open MCP config file and add Skavelo under mcpServers:

{
  "mcpServers": {
    "skavelo": { "serverUrl": "https://api.skavelo.com/mcp" }
  }
}

Save the file. Windsurf starts the sign-in for Skavelo by itself; approve the connection in Skavelo.

GitHub Copilot CLICopy Anchor Link

copilot plugin install Skavelo/ai-assistants

The plugin adds Skavelo's skills and the connection. To connect without the skills, run copilot mcp add --transport http skavelo https://api.skavelo.com/mcp instead. Start Copilot CLI and sign in when it asks; to start the sign-in yourself, run /mcp auth skavelo.

Other assistantsCopy Anchor Link

Any assistant that supports remote MCP servers with OAuth sign-in works, such as other editors and agents: add a remote (HTTP) server with the address above and start the sign-in from the assistant.

Skills for your assistantCopy Anchor Link

Skills are ready-made instructions that teach an assistant the common Skavelo jobs, step by step. In Claude Code, one plugin adds the skills and the connection to Skavelo:

/plugin marketplace add skavelo/ai-assistants
/plugin install skavelo@skavelo

In Cursor, the same plugin adds the skills and the connection: open Customize, choose From GitHub Repository, enter https://github.com/Skavelo/ai-assistants, then install skavelo.

In VS Code, Gemini CLI, the Antigravity CLI and GitHub Copilot CLI, the same repository installs the skills too; see their sections above. Other assistants that support skills in the SKILL.md format can use the same files from the skavelo/ai-assistants repository.

Things to askCopy Anchor Link

• "Write a walkthrough for sending an invoice, using what's in the product map."
• "Go through this week's Needs attention and suggest a note for each one."
• "Create lookup skills from our API spec at https://… and leave them switched off."
• "Turn our refund macro into a playbook, set to testing."
• "Hide the widget on the checkout pages."

Approving a connectionCopy Anchor Link

When the assistant connects, Skavelo shows which assistant is asking and where it will send you back. If you're signed out, you sign in first. Choose:

• Workspace, if you belong to more than one.
• Read and change, or Read only.

Only approve a connection you started yourself. The assistant then works as you, with your role: members can't change skills, playbooks or walkthrough offers through an assistant, the same as in the dashboard.

What it can doCopy Anchor Link

A few limits keep your users safe:

• Nothing can be deleted through an assistant.
• Some settings stay in the dashboard: test accounts for apps behind a login, the email address handed-off conversations go to, uploading files and the widget icon.
• Skills it creates start switched off. It can switch on a lookup, which only reads data. Actions change your users' data, so someone switches them on in the dashboard after checking them.
• New playbooks start off. Set one to testing to try it in Try it before your users see it.
• Everything goes through the same checks and plan limits as the dashboard.

Conversations and Needs attention include the names and emails your app passes to Skavelo, so an assistant with access can read them. Their questions were written by your users. Good assistants treat them as text to review, but read what yours proposes before you let it make changes.

Disconnect an assistantCopy Anchor Link

The AI assistants page in the dashboard lists every connected assistant, who connected it and when it was last used. Click the bin to disconnect one. Members can disconnect their own; owners and admins can disconnect anyone's. An assistant also stops working when the person who connected it leaves the workspace.

Scripts and CICopy Anchor Link

For a script with no person to approve it, send an API key instead: Authorization: Bearer sk_live_.... An API key can read and change everything the tools cover, like an admin.