Nimbus MCP Server

Connect your AI assistant to your workflow memory. This MCP server lets AI tools access your captured tasks and platform knowledge from Nimbus, so they can learn how you perform tasks and help you more effectively.

Quick Start

1. Get Your API Key

1. Open the Nimbus desktop app
2. Go to Settings → Account
3. Click Generate MCP API Key
4. Copy the key (starts with nimbus_)

2. Connect Your AI Tool

Choose your platform below.

Claude Desktop (Custom Connector — Recommended)

The easiest way to connect Nimbus to Claude Desktop:

1. Open Claude Desktop → Settings → Integrations → Add custom connector
2. Name: Nimbus
3. Remote MCP server URL: https://api.nimbusai.cloud/mcp
4. Leave OAuth Client ID and OAuth Client Secret blank (auto-discovered)
5. Click Add
6. Claude Desktop will open a browser window — paste your Nimbus API Key and click Authorize
7. Done! Nimbus tools will appear in Claude Desktop

Claude Desktop (JSON Config — Alternative)

If you prefer manual configuration, add to your config file:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "nimbus": {
      "command": "npx",
      "args": ["-y", "@nimbus-app/mcp"],
      "env": {
        "NIMBUS_API_KEY": "nimbus_your_key_here"
      }
    }
  }
}

Restart Claude Desktop (Cmd+Q / Ctrl+Q, reopen). You should see "nimbus" in the MCP tools list.

Claude Code (CLI)

One command:

claude mcp add nimbus -e NIMBUS_API_KEY=your_key -- npx -y @nimbus-app/mcp

To make it available across all projects:

claude mcp add --scope user nimbus -e NIMBUS_API_KEY=your_key -- npx -y @nimbus-app/mcp

Cursor

Add to your MCP config:

Global: ~/.cursor/mcp.json Per-project: .cursor/mcp.json in project root

{
  "mcpServers": {
    "nimbus": {
      "command": "npx",
      "args": ["-y", "@nimbus-app/mcp"],
      "env": {
        "NIMBUS_API_KEY": "nimbus_your_key_here"
      }
    }
  }
}

Windsurf

Add to your MCP config:

macOS/Linux: ~/.codeium/windsurf/mcp_config.json Windows: %USERPROFILE%\.codeium\windsurf\mcp_config.json

{
  "mcpServers": {
    "nimbus": {
      "command": "npx",
      "args": ["-y", "@nimbus-app/mcp"],
      "env": {
        "NIMBUS_API_KEY": "nimbus_your_key_here"
      }
    }
  }
}

OpenAI Codex CLI

One command:

codex mcp add nimbus --env NIMBUS_API_KEY=your_key -- npx -y @nimbus-app/mcp

Or add manually to ~/.codex/config.toml:

[mcp_servers.nimbus]
command = "npx"
args = ["-y", "@nimbus-app/mcp"]

[mcp_servers.nimbus.env]
NIMBUS_API_KEY = "nimbus_your_key_here"

Any MCP-Compatible Tool

The server works with any tool that supports the Model Context Protocol. Use stdio transport:

• Command: npx -y @nimbus-app/mcp
• Environment: NIMBUS_API_KEY=your_key
• Transport: stdio

Or install globally first:

npm install -g @nimbus-app/mcp

Then use nimbus-mcp as the command instead of npx -y @nimbus-app/mcp.

Available Tools

Once connected, your AI assistant has access to:

nimbus_search_tasks

Search for task patterns you've performed. Returns matching tasks with confidence scores and associated platforms.

Leave the query blank to browse your most recent analyzed tasks.

Example prompts:

• "Search my tasks for submitting homework on Canvas"
• "Find how I usually file expense reports"
• "Look for tasks involving VS Code"

nimbus_get_task

Get full details for a specific task, including platform knowledge and recent execution examples with step-by-step screenshots.

Example prompts:

• "Show me the full details for that Canvas task"
• "Get the execution steps with screenshots"

nimbus_search_workflows

Search for recorded workflows/sessions by natural language query. Returns matching workflows with status, counts, and top related tasks.

Example prompts:

• "Find workflows where I submitted homework"
• "Show my recent workflows about expense reports"

nimbus_get_workflow

Get a high-level overview of a recording session — which tasks the user performed, in what order, and the platforms involved.

Example prompts:

• "Show me what happened in that workflow"
• "Get the task breakdown for this session"

nimbus_search_platforms

Search for platforms (apps/websites) you work with. Returns full platform details including accumulated knowledge, identifiers, and related tasks.

Example prompts:

• "What platforms do I use for development?"
• "Find information about Canvas"

nimbus_ask_user

Ask you a question (you'll be notified in the Nimbus app).

nimbus_get_question_status

Check if you've answered a question.

nimbus_add_task_note

Leave a note on a task for future agents when you discover a workaround, gotcha, or non-obvious tip.

Example Conversation

You: Help me submit my CS101 homework on Canvas

Claude: Let me search your tasks to see how you usually do this.
[Uses nimbus_search_tasks with "submit homework Canvas"]

I found a task called "Submit Canvas Assignment" that you've done 5 times.
Let me get the detailed steps.
[Uses nimbus_get_task with include_executions=1]

Here's how you typically submit homework on Canvas:

1. Navigate to canvas.instructure.com
2. Click on "CS101" in the course list
3. Click "Assignments" in the left sidebar
4. Find your assignment and click on it
5. Click "Submit Assignment" button
6. Upload your file using the file picker
7. Click "Submit" to confirm

Would you like me to walk you through this step by step?

Configuration Options

Environment Variables

| Variable | Description | Default | |----------|-------------|---------| | NIMBUS_API_KEY | Your Nimbus API key (required) | - | | NIMBUS_API_URL | API endpoint URL | https://api.nimbusai.cloud/mcp/v1 |

Troubleshooting

"NIMBUS_API_KEY is required"

Make sure you've added the API key to your config and restarted the tool.

"Invalid or missing API key"

Your API key may be incorrect or expired. Generate a new one in Nimbus Settings → Account.

"Rate limit exceeded"

The API has limits of:

• 100 requests per minute
• 1,000 requests per hour
• 10,000 requests per day

Wait a moment and try again.

Server not appearing

1. Check your config file path is correct
2. Ensure the JSON is valid (no trailing commas)
3. Completely quit and restart the app
4. Try running manually: NIMBUS_API_KEY=your_key npx @nimbus-app/mcp

Node.js not found

You need Node.js 18+. Download from nodejs.org (LTS version).

Privacy & Security

• Your API key authenticates requests to your own workflow data
• Nimbus only captures what you've allowed in Privacy Settings
• Screenshots and workflow data are encrypted in transit and at rest
• You can delete your data anytime from the Nimbus app

Support

• Documentation: https://nimbusai.cloud/docs/mcp-setup
• Issues: https://github.com/nimbus-app/nimbus/issues
• Email: [email protected]

License

Proprietary. Copyright Nimbus. All rights reserved.