GnamiAI MCP connector

Read the chat and rendered app (v2.0.50+)

Call app_read with no arguments to request a fresh view. It returns commandId. Call app_read again with that commandId until status is executed. The result includes the active session, model, raw Markdown, rendered message text, headings, list items, code blocks and copy-button presence. A PNG of the current GnamiAI window is returned as MCP image content. The screenshot covers the current viewport; the structured messages include chat content outside it. Captures are processed in command order and retained for ten minutes. They are accessible only through tokens belonging to your account.

app_read({})
app_read({"commandId": "returned-id"})

Desktop command results (v2.0.49+)

app_control returns a queued command ID. Check app_status.commands for that ID: queued, delivered, executed, failed, or expired. Delivered means the app received it; executed means its handler finished. A prompt can wait for your existing in-app permission approval. Errors are returned with failed commands. Commands expire after five minutes without acknowledgement; queue fresh commands after updating.

app_status also returns version, lastPollAt, pending, and connectedAt. Use set_model with an exact model ID from the app, then send_prompt. Commands are processed in order. Billing data is limited to your account.

Hosted endpoint

API authentication

Sign in at Account, create an MCP API token, and copy it when shown. The token is stored as a hash and cannot be viewed again.

Available tools

• subscription_status — your own subscription tier, limits, and subscription records only.
• app_status — whether your signed-in desktop is connected.
• app_control — queue open_mode, open_agent, new_session, or send_prompt for your desktop.

Access requirements

• A signed-in GnamiAI customer account is required to create a token.
• The desktop app must be signed in to the same customer account and running for remote commands.
• MCP is strictly customer-scoped: it does not expose revenue, other customers, or aggregate subscription data.

MCP client config

{
  "mcpServers": {
    "gnamiai": {
      "url": "https://gnamiai.com/api/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_GNAMIAI_MCP_TOKEN"
      }
    }
  }
}

VS Code

In VS Code, run MCP: Open User Configuration and add the HTTP server below. Replace the token with one created in your GnamiAI account. The tools appear in VS Code's agent chat and control your running GnamiAI desktop; the native VSIX above handles project handoff.

{
  "servers": {
    "gnamiai": {
      "type": "http",
      "url": "https://gnamiai.com/api/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_GNAMIAI_MCP_TOKEN"
      }
    }
  }
}

JetBrains IDEs

In an IDE with JetBrains AI Assistant, open Settings > Tools > AI Assistant > Model Context Protocol (MCP), choose Add > HTTP, and enter this configuration with your account token. The connector exposes GnamiAI desktop tools inside AI Assistant; the native plugin above handles project handoff.

{
  "mcpServers": {
    "gnamiai": {
      "type": "streamable-http",
      "url": "https://gnamiai.com/api/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_GNAMIAI_MCP_TOKEN"
      }
    }
  }
}