OrbitAI

Your AI-powered MongoDB Atlas mission control.

OrbitAI is a TypeScript monorepo providing conversational access to MongoDB Atlas through an MCP server and AI terminal shell. It enables platform teams, DevOps engineers, and administrators to manage Atlas infrastructure using natural language.

Architecture

┌─────────────────────────────────────────────────────────────────┐
│                      MCP Server                                 │
│                                                                 │
│  ┌─────────────────┐      ┌──────────────────────────────────┐  │
│  │ StdioTransport  │      │ HTTP Server (:3600)              │  │
│  │ (Claude Desktop)│      │ StreamableHTTPServerTransport    │  │
│  └────────┬────────┘      └──────────────┬───────────────────┘  │
│           │                              │                      │
│           └──────────────┬───────────────┘                      │
│                          ▼                                      │
│                   ┌─────────────┐                               │
│                   │   Server    │                               │
│                   │  67 Tools   │──── @orbit/core               │
│                   │ 15 Resources│        │                      │
│                   │  5 Prompts  │        ▼                      │
│                   └─────────────┘   Atlas Admin API             │
└─────────────────────────────────────────────────────────────────┘

┌─────────────────────────────────────────────────────────────────┐
│                        orbit-ai CLI                             │
│                                                                 │
│  ┌───────────────────────────────────────────────────────────┐  │
│  │  McpClientWrapper                                         │  |
│  │  ├─ Primary: HTTP (http://127.0.0.1:3600/mcp)             │  │
│  │  └─ Fallback: Stdio (spawn orbit-mcp-server)              │  │
│  └───────────────────────────────────────────────────────────┘  │
│                          │                                      │
│                          ▼                                      │
│                 ┌────────────────┐                              │
│                 │   Agent Loop   │                              │
│                 │   (4 LLMs)     │                              │
│                 └────────────────┘                              │
└─────────────────────────────────────────────────────────────────┘

Packages

Package	Description
`@orbit/core`	Atlas API client, HTTP Digest auth, unified configuration, 41 domain action maps
`@orbit/mcp-server`	MCP server with HTTP and stdio transports, 67 tools, 15 resources, 5 prompts
`@orbit/cli`	`orbit-ai` — conversational terminal shell with Anthropic, OpenAI, Google, and Ollama support

Capabilities

Cluster Management — Create, scale, pause, resume, and configure dedicated, Flex, serverless, and global clusters
Database Tools — Connect to MongoDB instances, query collections, inspect schemas, run aggregations
Security & Access — Database users, custom roles, IP access lists, network peering, private endpoints
Monitoring & Alerts — Real-time metrics, performance advisor, slow query analysis, alert configurations
Backup & Recovery — Cloud backup snapshots, point-in-time restore, export to cloud storage
Atlas Services — Search indexes, Vector Search, Data Federation, Stream Processing
Billing & Cost — Invoices, cost explorer, line items

Prerequisites

Node.js 18+
npm 9+
MongoDB Atlas account with API keys

Quick Start

1. Clone and Build

git clone https://github.com/corbtastik/orbit.git
cd orbit
npm install
npm run build

2. Configure

Create ~/.orbit-ai/config.json:

{
  "atlas": {
    "publicKey": "your-atlas-public-key",
    "privateKey": "your-atlas-private-key"
  },
  "llm": {
    "provider": "anthropic"
  }
}

Set your LLM API key:

export ANTHROPIC_API_KEY="your-key"

3. Run

Option A: CLI with stdio (single process)

node apps/cli/dist/index.js

Option B: CLI with HTTP server (recommended for development)

Terminal 1 — Start the MCP server:

ORBIT_MCP_HTTP=true node packages/mcp-server/dist/index.js

Terminal 2 — Start the CLI:

node apps/cli/dist/index.js

The CLI auto-connects via HTTP when the server is running.

Configuration

Configuration priority: CLI flags > Environment variables > Config file > Defaults

Config File

Location: ~/.orbit-ai/config.json

{
  "atlas": {
    "publicKey": "your-public-key",
    "privateKey": "your-private-key",
    "orgId": "default-org-id",
    "groupId": "default-project-id",
    "baseUrl": "https://cloud.mongodb.com"
  },
  "llm": {
    "provider": "anthropic",
    "model": "claude-sonnet-4-20250514",
    "apiKey": "your-api-key",
    "baseUrl": null,
    "maxTokens": 4096,
    "temperature": 0
  },
  "mcp": {
    "url": "http://127.0.0.1:3600/mcp",
    "forceStdio": false,
    "httpTimeout": 2000
  },
  "mongodb": {
    "connections": {
      "local": "mongodb://localhost:27017",
      "atlas": "mongodb+srv://user:pass@cluster.mongodb.net"
    }
  },
  "defaults": {
    "outputFormat": "text",
    "maxToolTurns": 10,
    "verbose": false
  }
}

Environment Variables

Atlas API

Variable	Description	Default
`ATLAS_PUBLIC_KEY`	MongoDB Atlas public API key (required)	—
`ATLAS_PRIVATE_KEY`	MongoDB Atlas private API key (required)	—
`ATLAS_ORG_ID`	Default organization ID	—
`ATLAS_GROUP_ID`	Default project/group ID	—
`ATLAS_BASE_URL`	Atlas API base URL	`https://cloud.mongodb.com`

LLM Provider

Variable	Description	Default
`ORBIT_LLM_PROVIDER`	Provider: `anthropic`, `openai`, `google`, `ollama`	`anthropic`
`ORBIT_LLM_MODEL`	Model name	Per-provider default
`ORBIT_LLM_API_KEY`	API key (overrides provider-specific keys)	—
`ORBIT_LLM_BASE_URL`	Custom API base URL	—
`ORBIT_LLM_MAX_TOKENS`	Max response tokens	`4096`
`ORBIT_LLM_TEMPERATURE`	Temperature	`0`

Provider-Specific API Keys

Variable	Provider
`ANTHROPIC_API_KEY`	Anthropic (Claude)
`OPENAI_API_KEY`	OpenAI (GPT)
`GOOGLE_API_KEY`	Google (Gemini)

MCP Configuration

Variable	Description	Default
`ORBIT_MCP_HTTP`	Start MCP server in HTTP mode	`false`
`ORBIT_MCP_PORT`	HTTP server port	`3600`
`ORBIT_MCP_HOST`	HTTP server host	`127.0.0.1`
`ORBIT_MCP_URL`	MCP server URL (for CLI)	`http://127.0.0.1:3600/mcp`
`ORBIT_MCP_STDIO`	Force CLI to use stdio transport	`false`
`ORBIT_MCP_HTTP_TIMEOUT`	HTTP connection timeout (ms)	`2000`

MongoDB Connections

Define named connections using the MONGODB_CONN_<NAME> pattern:

export MONGODB_CONN_LOCAL="mongodb://localhost:27017"
export MONGODB_CONN_ATLAS="mongodb+srv://user:pass@cluster.mongodb.net"
export MONGODB_CONN_PROD="mongodb+srv://user:pass@prod.mongodb.net"

The <NAME> becomes the connection identifier used in tool calls.

Running the MCP Server

Stdio Mode (Default)

Used by Claude Desktop, Cursor, and other MCP clients:

node packages/mcp-server/dist/index.js

HTTP Mode

For development and the orbit-ai CLI:

ORBIT_MCP_HTTP=true node packages/mcp-server/dist/index.js

Or with flags:

node packages/mcp-server/dist/index.js --http --port 3600

Claude Desktop Configuration

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "orbit": {
      "command": "node",
      "args": ["/path/to/orbit/packages/mcp-server/dist/index.js"],
      "env": {
        "ATLAS_PUBLIC_KEY": "your-public-key",
        "ATLAS_PRIVATE_KEY": "your-private-key",
        "MONGODB_CONN_LOCAL": "mongodb://localhost:27017",
        "MONGODB_CONN_ATLAS": "mongodb+srv://user:pass@cluster.mongodb.net"
      }
    }
  }
}

Running the CLI

Interactive Mode

node apps/cli/dist/index.js

One-Shot Mode

node apps/cli/dist/index.js "list my clusters"

CLI Options

-p, --provider <name>   LLM provider: anthropic, openai, google, ollama
-m, --model <model>     Model name
    --api-key <key>     API key for the LLM provider
-v, --verbose           Show tool details and token usage
    --max-tokens <n>    Max response tokens (default: 4096)
    --raw               Output raw markdown (disable rendering)
-h, --help              Show help
    --version           Show version

Shell Commands

Command	Description
`/help`	Show available commands
`/config`	Show current configuration
`/clear`	Clear conversation history
`/quit`	Exit the shell

Example Session

  ◉ orbit-ai v0.2.0
  anthropic/claude-sonnet-4-20250514

❯ list my projects
  ✦ manage_projects → list

  You have 3 projects:
  - production (64a1b2c3d4e5f6a7b8c9d0e1)
  - staging (64a1b2c3d4e5f6a7b8c9d0e2)
  - dev (64a1b2c3d4e5f6a7b8c9d0e3)

❯ summarize the mongoball database schema
  ✦ list-collections → unknown
  ✦ collection-schema → unknown

  The mongoball database contains 5 collections:
  - people: Player biographical information
  - teams: Yearly team statistics
  - batting: Player batting statistics by year
  ...

❯ /quit

Development

Build

# Build all packages
npm run build

# Build a single package
npm run build -w @orbit/core
npm run build -w @orbit/mcp-server
npm run build -w @orbit/cli

Test

# Run all tests (228 tests)
npm test

# Run tests for a single package
npm test -w @orbit/core

Clean

npm run clean

Project Structure

orbit/
├── packages/
│   ├── core/              @orbit/core
│   │   └── src/
│   │       ├── auth/          Digest auth
│   │       ├── client/        AtlasClient
│   │       ├── config/        Unified config loader
│   │       ├── domains/       41 domain action maps
│   │       └── errors/        Error types
│   └── mcp-server/        @orbit/mcp-server
│       └── src/
│           ├── tools/         67 tools (Atlas + Database)
│           ├── transport/     HTTP and stdio transports
│           ├── resources.ts   15 resources
│           └── prompts.ts     5 prompts
├── apps/
│   └── cli/               @orbit/cli
│       └── src/
│           ├── agent/         Conversation loop
│           ├── config/        CLI config
│           ├── mcp/           MCP client wrapper
│           ├── providers/     LLM providers
│           └── ui/            Terminal UI
└── docs/                  Documentation

Using Local LLMs

Ollama

export ORBIT_LLM_PROVIDER="ollama"
export ORBIT_LLM_MODEL="llama3.1"
export ORBIT_LLM_BASE_URL="http://localhost:11434/v1"

node apps/cli/dist/index.js

LM Studio (OpenAI-compatible)

export ORBIT_LLM_PROVIDER="openai"
export ORBIT_LLM_MODEL="qwen2.5-coder-14b"
export ORBIT_LLM_BASE_URL="http://localhost:1234/v1"
export ORBIT_LLM_API_KEY="lm-studio"

node apps/cli/dist/index.js

License

MIT

Author

@corbtastik

Name		Name	Last commit message	Last commit date
Latest commit History 35 Commits
apps		apps
demo		demo
docs		docs
packages		packages
scripts		scripts
.gitignore		.gitignore
README.md		README.md
package-lock.json		package-lock.json
package.json		package.json
tsconfig.base.json		tsconfig.base.json
vitest.workspace.ts		vitest.workspace.ts

Folders and files

Latest commit

History

Repository files navigation

OrbitAI

Architecture

Packages

Capabilities

Prerequisites

Quick Start

1. Clone and Build

2. Configure

3. Run

Configuration

Config File

Environment Variables

Atlas API

LLM Provider

Provider-Specific API Keys

MCP Configuration

MongoDB Connections

Running the MCP Server

Stdio Mode (Default)

HTTP Mode

Claude Desktop Configuration

Running the CLI

Interactive Mode

One-Shot Mode

CLI Options

Shell Commands

Example Session

Development

Build

Test

Clean

Project Structure

Using Local LLMs

Ollama

LM Studio (OpenAI-compatible)

License

Author

About

Topics

Resources

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages