Keep Claude Desktop.
Give it your machines.
AI Commander is not an alternative to Claude Desktop, Claude Code, Codex, ChatGPT, Cursor, GitHub Copilot, Sourcegraph Cody, Grok, Gemini, Windsurf, Continue, Cline, Roo Code, Zed, OpenCode, Qwen Code, goose, Amp, Kilo Code, OpenHands, Amazon Q Developer, Augment Code, Devin, JetBrains Junie, Kiro, Warp, Factory, Crush, LM Studio, Trae, n8n, or OpenAI Agents SDK. It is the remote-machine harness those tools plug into over MCP: outbound-only shell, detached jobs that outlive the chat, and file transfer on hardware you already own.
Sibling harness pages: Claude Code · Codex · ChatGPT · Cursor · OpenCode · Factory · Kiro · Warp · Crush · LM Studio · Trae · n8n · OpenAI Agents SDK · Gemini — and the rest on Connect your AI client.
Connect Claude Desktop (custom connector)
Claude Desktop adds remote MCP servers as custom connectors (Claude Help Center: custom connectors using remote MCP). On an individual Pro or Max plan, go to Customize → Connectors, click + Add → Add custom connector, give it a name, and paste AI Commander’s hosted endpoint:
Name: AI Commander URL: https://aicommander.dev/mcp
Click Continue and review the authentication Claude detected. AI Commander supports OAuth with Dynamic Client Registration, so if Claude asks how to identify itself, Register automatically works with no credentials to paste; you then finish a one-time browser sign-in that binds the connector to your account, so saved-machine aliases resolve. Prefer pre-provisioned credentials? Create an OAuth client in the AI Commander dashboard and choose Use your own OAuth client (details on Connect your AI client). On Team and Enterprise plans an Owner adds the connector under Organization settings → Connectors first, then each member connects it. Free plans are limited to one custom connector.
Then enable it per conversation from the + button → Connectors toggle. Older Claude Desktop builds label the menu Settings → Connectors — check the linked Help Center article for your version.
Why an outbound-only relay fits custom connectors
Anthropic’s docs are explicit: a custom connector is reached from Anthropic’s cloud, not from your laptop, even in Claude Desktop.
Servers on a private network, behind a VPN, or a firewall will not connect unless you allowlist Anthropic’s IP ranges.
AI Commander sidesteps that: the connector talks to the public relay at https://aicommander.dev/mcp, and each of your machines
keeps an outbound-only connection to the relay — no inbound ports, no exposed SSH, nothing to allowlist on the box.
Prefer a local stdio server instead? Claude Desktop also reads claude_desktop_config.json
(macOS: ~/Library/Application Support/Claude/, Windows: %APPDATA%\Claude\; or Settings → Developer → Edit Config):
{
"mcpServers": {
"aicommander": {
"command": "npx",
"args": [
"-y",
"@aicommander/mcp"
],
"env": {
"AICOMMANDER_SERVER": "https://aicommander.dev"
}
}
}
}
Restart Claude Desktop after editing the file. On Windows use "command": "npx.cmd", and add
"AICOMMANDER_TOKEN": "<your-key>" to env for alias-based access.
Local servers from this file are separate from connectors and are not available on claude.ai.
Keep Claude’s tool approval prompts on for AI Commander at first, and only use “Allow always” for tools you trust to run unsupervised;
long builds and training runs are good candidates for detached jobs (remote_job_start) so the chat does not have to wait.
For a code-only first hour without signing in, use https://aicommander.dev/mcp?anonymous=1 as the connector URL.
Using Claude Code in the terminal as well? See Keep Claude Code. Full client matrix (Claude Desktop, Claude Code, n8n, OpenAI Agents SDK, Codex, ChatGPT, Cursor, Copilot, Gemini, Trae, Crush, LM Studio, and more, plus REST): Connect your AI client.
What you get that the chat alone cannot
- Remote shell on Linux, macOS, and Windows — by session code or saved alias.
- Detached jobs for builds, test suites, and training on the box that has the repo.
- Fleet list + GPU facts so Claude picks the right machine before it burns an hour.
- File courier (Pro) for checkpoints and diffs up to 100 MiB.
- Outbound-only agent on each machine — no inbound ports, no exposed SSH, no VPN to set up.
Not a replacement — a multiplier
| You already use | AI Commander adds |
|---|---|
| Claude Desktop | Customize → Connectors → Add custom connector → https://aicommander.dev/mcp |
| n8n | AI Agent → Tool → MCP Client Tool, Endpoint https://aicommander.dev/mcp, HTTP Streamable |
| OpenAI Agents SDK | MCPServerStreamableHttp(params={"url": "https://aicommander.dev/mcp", …}) or HostedMCPTool |
| Claude Code | claude mcp add --transport http … → remote exec + jobs |
| Codex / ChatGPT | Same MCP endpoint or REST /api/v1/exec |
| Cursor / Continue / Windsurf | npx @aicommander/mcp stdio bridge |
FAQ
https://aicommander.dev/mcp as the remote MCP server URL. AI Commander supports Dynamic Client Registration, so Register automatically needs no credentials; finish the one-time browser sign-in to bind the connector to your account. Prefer a local server? Put mcpServers.aicommander with command npx and args -y @aicommander/mcp in claude_desktop_config.json and restart. Use ?anonymous=1 for a code-only first hour. See /howto/ for the full matrix.https://aicommander.dev/mcp. Your machines only keep an outbound-only connection to that relay — no inbound ports, no exposed SSH, no VPN, and no IP allowlist on the machine.claude mcp add --transport http …). This page covers the Claude Desktop app: custom connectors and claude_desktop_config.json. Both reach the same AI Commander machines.Next steps
Install the agent on a machine you control, add AI Commander as a Claude Desktop custom connector (https://aicommander.dev/mcp), then ask Claude for a real task on a remote box.