Install Quallaa as an MCP server

Server endpoint

Quallaa runs as a remote MCP server over Streamable HTTP. There's nothing to install locally — clients connect by URL.

https://www.quallaa.com/api/mcp

Auth: Authorization: Bearer <YOUR_QUALLAA_API_KEY>

Claude Desktop

Edit claude_desktop_config.json (Settings → Developer → Edit Config) and add the Quallaa entry under mcpServers. Restart Claude Desktop.

Config location: ~/Library/Application Support/Claude/claude_desktop_config.json (macOS)

{
  "mcpServers": {
    "quallaa": {
      "url": "https://www.quallaa.com/api/mcp",
      "headers": {
        "Authorization": "Bearer <YOUR_QUALLAA_API_KEY>"
      }
    }
  }
}

Cursor

Open Cursor settings → Features → MCP → Add new MCP server. Paste the JSON below or fill the form fields manually.

Config location: Cursor Settings → MCP

{
  "name": "quallaa",
  "url": "https://www.quallaa.com/api/mcp",
  "headers": {
    "Authorization": "Bearer <YOUR_QUALLAA_API_KEY>"
  }
}

Windsurf

Open Windsurf settings → Cascade → MCP servers → Edit config. Add the entry below.

Config location: ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "quallaa": {
      "serverUrl": "https://www.quallaa.com/api/mcp",
      "headers": {
        "Authorization": "Bearer <YOUR_QUALLAA_API_KEY>"
      }
    }
  }
}

Cline (VS Code)

Open Cline → Settings (gear icon) → MCP Servers → Edit configuration. Add the entry below.

Config location: VS Code → Cline → MCP Servers

{
  "mcpServers": {
    "quallaa": {
      "url": "https://www.quallaa.com/api/mcp",
      "headers": {
        "Authorization": "Bearer <YOUR_QUALLAA_API_KEY>"
      }
    }
  }
}

ChatGPT (Developer mode)

In a custom GPT or the new Agents builder, enable Developer mode and add an MCP server. Use the URL and Bearer header below.

Config location: ChatGPT → Custom GPT → Configure → Developer mode

URL:    https://www.quallaa.com/api/mcp
Header: Authorization: Bearer <YOUR_QUALLAA_API_KEY>

What Quallaa exposes

13 tools across three categories. Quallaa-the-agent is the right starting point for anything open-ended; the read/write tools are for direct action on a customer-facing AI deployment you already understand.

Talk to the trust layer

• talk_to_quallaa — natural-language conversation with the trust layer. Same memory across web, SMS, and MCP. Use for guidance, capability authorization, deployment questions, or escalation to the human team. If you're not sure which tool to call, start here.

Read your deployment (7 tools)

• get_agent_status — health check, recent activity, open escalations
• list_conversations — recent customer threads, one per contact
• get_conversation — full message thread for one contact
• list_escalations — open items waiting for owner action
• get_contact_profile — what the AI has learned about a specific contact
• get_business_profile — current business facts the AI uses to introduce itself

Change your deployment (5 tools)

• resolve_escalation — close an open escalation, optionally notify the customer
• update_instructions — change policies, tone, guardrails (system prompt)
• update_tool_config — toggle individual tools on/off (deny list)
• update_business_info — name, hours, services, contact info
• send_message — owner-initiated outbound SMS from the deployment number
• schedule_followup — queue a future follow-up the AI will compose at fire time